feat(common): add IConfiguration.VendorExtensions for XSD extension

[ottobolyos]

Summary

Adds a standard-compliant vendor-extension surface on IConfiguration — a new IEnumerable<XElement> VendorExtensions property — that carries fully-formed, vendor-namespaced XML elements verbatim through the MTConnect.NET XML formatter. The mechanism mirrors the MTConnect v2.7 XSD's own extension design: every standard child of ComponentConfigurationType is a substitutionGroup='AbstractConfiguration' declaration, and vendors extend the type by publishing their own XSD that likewise substitutes for AbstractConfiguration. VendorExtensions is the strongly-typed slot that carries those substitutions end-to-end.

Standard citations

MTConnectDevices_2.7.xsd:

• ComponentConfigurationType (line 8161) declares its content model as <xs:element ref="AbstractConfiguration" minOccurs="1" maxOccurs="unbounded"/> — the sequence permits any element that substitutes for the abstract AbstractConfiguration element and rejects unknown children otherwise (no xs:any).
• Substitution declarations (lines 8239 through 9974) mark every standard child — SensorConfiguration, Specifications, Relationships, CoordinateSystems, Motion, SolidModel, ImageFiles, PowerSources — with substitutionGroup='AbstractConfiguration'. Vendors extend the type by publishing an XSD declaring a vendor-namespaced element with the same substitutionGroup.

The MTConnect.NET XML formatter writes each VendorExtensions entry verbatim inside the <Configuration> sequence via XmlWriter.WriteRaw. Downstream strict-XSD validation continues to work when the vendor's XSD is loaded alongside the MTConnect schemas — the vendor's XSD supplies the substitution declaration that makes the element valid at the AbstractConfiguration slot.

What this PR does

• feat(sysml-import) — marks Devices.Configurations.Configuration as IsPartial = true in the CSharp renderer, matching the pattern already applied to Component, Composition, DataItem, and other classes that carry hand-authored companion partials.
• feat(common) — regenerates Configuration.g.cs and IConfiguration.g.cs from the v2.7 XMI to declare partial class / partial interface.
• feat(common) — adds hand-authored Configuration.cs and IConfiguration.cs partials that declare IEnumerable<XElement> VendorExtensions { get; } on the interface and { get; set; } on the concrete.
• feat(xml) — wires the surface through XmlConfiguration: [XmlAnyElement] capture on the read path, WriteRaw emission on the write path, ordered after the standard children.

Non-goals

This is not a raw-string RawXml passthrough — a string-typed slot bypasses namespace hygiene and would produce documents that fail strict XSD validation at the ComponentConfigurationType sequence. The XElement typing forces well-formed XML at author time and preserves namespace declarations verbatim through the formatter.

Breaking

Adds a new member (VendorExtensions { get; }) to the public IConfiguration interface — a source-compatibility break for any external implementer of IConfiguration outside the library. Acceptable on the next major version boundary.

---

Supersedes #214.

Dime review cycle 1

Retroactive backfill (2026-08-20). The initial 6-agent Ultrareview cycle exercised the XML round-trip surface end-to-end. Ledger reconstruction from commit history:

• [TEST] test-coverage-audit — XML round-trip pin for Configuration.VendorExtensions (empty, single-entry, multi-entry) landed as test(xml-tests): pin XML round trip for Configuration.VendorExtensions (d3d0d88).
• [TEST] test-coverage-audit — null-element handling on the interface surface pinned by test(xml-tests): pin null-element handling on IConfiguration.VendorExtensions (fe095bf).
• [TEST] test-coverage-audit — HTTP Probe end-to-end coverage added via test(integration): pin HTTP Probe end-to-end for IConfiguration.VendorExtensions (e8582e8).
• [TEST] test-coverage-audit — coverage widening for the interface implementation matrix added in test(xml-tests): widen IConfiguration.VendorExtensions coverage on interface (fe18487).

Dime review cycle 2

Retroactive backfill (2026-08-20). Second cycle after the 2026-08-19 BrE→AmE message sweep reconciliation, dispatched to close coverage FLOOR gaps surfaced by re-running test-coverage-audit on the reconciled tip. Ledger:

• [TEST] test-coverage-audit — coverage FLOOR gaps on IConfiguration.VendorExtensions (5 test files, +1165 LOC) closed atomically in test(xml-tests): close coverage FLOOR gaps on VendorExtensions (611f482). This commit was cherry-picked on top of the new AmE tip after the BrE→AmE swept in patch content and message (behaviour→behavior, serialise→serialize, honour→honor).
• Verified via nohup test on bluefin with --blame-hang-timeout 15min: all individual test suites green (Common 3987, XML 131, JSON 68, MqttRelay 62, Docs 67 = 4315 tests, 0 failed). Test host crashed at 15-min inactivity timeout during concurrent Docs-Tests execution — bluefin-flake pattern documented in fix: clean every Release-pack warning and add multi-TFM CI gate #219 + fix(xml): dispatch v2.6/v2.7 namespaces + default unknown to Max #229 verify reports (docs-tests hang under concurrent bluefin load, uncorrelated with PR content since PR doesn't touch docs surface).

(Zero unfixed findings — Ready-eligible.)

Dime review cycle 3

6-agent Ultrareview pass (2026-08-25) ahead of the Testing integration → Completed flip: bug-detector, code-review, security-audit, test-coverage-audit, documentation-audit, simplification.

• [FIXED] security-audit MEDIUM — XElement.Parse on each captured vendor-extension element's OuterXml used implicit XmlReader defaults instead of the repo's explicit XXE-hardening idiom (XmiDeserializer.FromXml/FromFile: pinned DtdProcessing.Prohibit + XmlResolver = null). Not currently exploitable — a DOCTYPE cannot legally appear inside a captured element's OuterXml — but closed for defense-in-depth consistency. fix(xml): harden VendorExtensions XElement parse against XXE (13c9af7).
• [FIXED] code-review MEDIUM — class remarks on IConfiguration.VendorExtensions misquoted ComponentConfigurationType's AbstractConfiguration ref as minOccurs="0"; the schema (MTConnectDevices_2.7.xsd:8168) declares minOccurs='1'. Corrected in both this PR description (above) and the doc comment. fix(common,xml-tests): correct XSD minOccurs citation, sweep AmE spelling (f0d1147).
• [FIXED] code-review LOW — BrE spellings (unrecognised, behaviour, serialise(s/d), deserialiser) introduced by this PR's own commits, including two test method names, swept to AmE in the same commit as above.
• [SKIP-rationale] security-audit MEDIUM — no depth/size ceiling on vendor-extension capture. DTD is already prohibited on both the outer document read and (after the fix above) the per-element re-parse, so no entity-expansion amplification vector exists; a large-but-linear nested payload is bounded by the same document-size limits as any other XML content in the ingestion pipeline, and no other slot on XmlConfiguration enforces a bespoke depth/size cap either. Not an inconsistency worth a one-off guard here.
• [SKIP-rationale] security-audit LOW — vendor extension's own namespace-prefix declarations could locally shadow a prefix already open in the writer's ancestor scope. XElement.ToString() always emits well-formed, self-contained XML (not a raw-string injection risk); a non-namespace-aware downstream consumer choosing to key off prefix text rather than the resolved namespace URI is off-spec on its own terms.
• [TRACKED] bug-detector MEDIUM — Configuration.ImageFiles/Configuration.PowerSource (both cited as standard AbstractConfiguration substitution children in this PR's own "Standard citations" section) have no XmlConfiguration wiring at all — pre-existing, repo-wide gap (confirmed absent on upstream/master before this PR; no XmlImageFile/XmlPowerSource surrogate exists anywhere in the tree). Before this PR such elements were silently dropped on read; after this PR's [XmlAnyElement] catch-all, they're captured into VendorExtensions instead — still functionally absent from the typed properties, but now mislabeled as vendor content. Filed as XML wire format missing for Configuration.ImageFiles / Configuration.PowerSource #262 rather than fixed in-PR: closing it properly means adding two new XML surrogate type families from scratch, a distinct feature from the VendorExtensions mechanism this PR delivers.
• [SKIP-rationale] bug-detector LOW — vendor-extension write order doesn't preserve original document-order interleave with standard children (always appended after Specifications). Consistent with every other slot on XmlConfiguration, which already serializes each standard child in a fixed field order regardless of source order — not a regression this PR introduces.
• [SKIP-rationale] bug-detector LOW — WriteRaw bypasses the writer's indent-tracking state, so raw-inserted extension content can look inconsistently indented relative to siblings when Settings.Indent = true. Cosmetic only.
• [SKIP-rationale] documentation-audit LOW — no VitePress page documents <Configuration>'s per-element wire format at all (pre-existing gap predating this PR; docs/wire-formats/xml.md is envelope-level, not per-element, for every other Configuration child too).
• [SKIP-rationale] simplification LOW × 2 — test-fixture scaffolding duplicated against sibling ConfigurationPolymorphicHttpProbeWorkflowTests.cs, and structurally parallel JSON/JSON-cppagent known-limit test files. Both consistent with this repo's existing one-fixture-per-file convention for independently-movable test files; not accidental duplication.
• [CLEAN] test-coverage-audit — coverage substantively complete (enum-arm N/A, all failure/branch paths covered, no test smells). Noted a TDD-ordering process gap (feat commits landed before test commits in cycle 1) — historical, not a content gap, not re-litigated.

All local verification re-run green after the fix commits: dotnet build -warnaserror 0/0, dotnet format --verify-no-changes clean, MTConnect.NET-XML-Tests 131/131, MTConnect.NET-Common-Tests 4102/4102, MTConnect.NET-HTTP-Tests 112/112.

(Zero unfixed MEDIUM+ findings — Completed-eligible.)

Depends on

• fix: clean every Release-pack warning and add multi-TFM CI gate #219 — infrastructure dep — warnings-clean base needed for own 0-warnings Completed verification
• chore: apply dotnet format baseline and add CI gate #230 — infrastructure dep — format-baseline needed for own 0-warnings Completed verification