Bind scalar ABI rows to source claims

[jscott3201]

Summary

• add a frozen internal projection that joins named scalar ABI rows with inventory-anchored caller source claims
• use named rows as the ordering and ABI authority while matching claims by scalar name within each namespace
• preserve detached primitive, alias, enum, value, direction, coordinate, and source-locator payload
• reject divergent metadata, namespaces, owners, coordinates, and malformed nested values before allocating output

The new projection is an in-memory lowering input. Source class and member values remain caller claims rather than verified declarations. This change adds no serialization, graph topology, CXF, deployment identity, registry content, public schema, or production routine.

Validation

• python3 -m unittest tools.lint.tests.test_routine_resolution tools.lint.tests.test_routine_scalar_abi tools.lint.tests.test_routine_scalar_names tools.lint.tests.test_routine_scalar_source_claims tools.lint.tests.test_routine_bound_scalars -v (74 tests)
• python3 -m py_compile tools/lint/routine_bound_scalars.py tools/lint/tests/test_routine_bound_scalars.py
• python3 -m unittest discover -s tools/lint/tests -v (179 tests, pinned schema environment)
• python3 tools/lint/routine_schemas.py
• python3 tools/lint/routine_semantics.py
• python3 tools/lint/routines.py
• cargo test --manifest-path tools/verify/Cargo.toml (11 tests)
• cargo run --quiet --manifest-path tools/verify/Cargo.toml -- --routines
• cargo run --quiet --manifest-path tools/verify/Cargo.toml -- --all
• python3 -m unittest discover -s tools/dataset_harness/tests -v (32 tests)
• exact tiny LBL FPU replay and assertions from verify.yml
• python3 tools/book/generate.py && mdbook build book with no tracked generated changes
• git diff --check

The exact two-checkout G36 inventory command was not rerun locally because the pinned source checkouts were unavailable. The base commit passed that gate, and pull-request CI runs it against both exact pins.

[jscott3201]

Immutable review cycle 1 targeted 40abea8de02fb09dc881b3920e35abde06c988e1.

Both standard review lenses are blocker/major-clean. No PR-introduced findings were deferred. The verify and book checks completed successfully on the reviewed head.