test(executors): restore the #10986 reasoning-only fallback guards

[pacocartones]

Summary

• Restores the two regression tests that [BUG] OpenAI API returns null content when model output consists of reasoning tokens (e.g. command-code/meta/muse-spark-1.2-contributor) #10986 added for reasoning-only Command Code output. The feat(providers): use official provider API (/provider/v1) instead of CLI-only /alpha/generate for command-code chat #10265 rewrite of tests/unit/command-code-executor.test.ts (b6412c6) deleted them, even though the production fallback in createJsonResponse / createStreamResponse (open-sse/executors/commandCode.ts) survived that rewrite, leaving it unguarded.
• Since feat(providers): use official provider API (/provider/v1) instead of CLI-only /alpha/generate for command-code chat #10265 the CLI translator is only reachable through the /alpha/generate fallback taken when /provider/v1/chat/completions answers 403/404, so both tests are routed through that path via a shared goPlanFallbackFetch() helper.
• Non-stream: a reasoning-only /alpha/generate stream yields message.content equal to the reasoning text, reasoning_content stays populated, reasoning_tokens is carried through from totalUsage.
• Stream: exactly one synthesized content delta is emitted, after the reasoning delta and before the finish chunk.

Related Issues

• Related to [BUG] OpenAI API returns null content when model output consists of reasoning tokens (e.g. command-code/meta/muse-spark-1.2-contributor) #10986
• Related to fix(backend): OpenAI API returns null content when model output consists of reasoning tokens (e.g. command-code/meta/muse-spark-1.2-contributor) #12130

Validation

• Change type: other (test-only hardening)
• Focused tests and category gates from the golden path: tests/unit/command-code-executor.test.ts via the repository runner, 18 pass / 0 fail. Disabling the three fallback guards in commandCode.ts makes exactly these two tests fail (non-stream content becomes ""; stream emits 0 content deltas); the file was restored unchanged afterward.
• npm run lint — repository-wide eslint exit 0 (run with --pass-on-unpruned-suppressions; the literal command reports only pre-existing unused global suppressions on this base); npm run typecheck:core 0; git diff --check clean; npm run check:changelog-integrity OK
• Reconciled with the current active release base release/v3.8.51@158647618; focused checks rerun afterward
• Production-code changes include a new or updated automated test in this PR: not applicable, no production code changed
• SonarQube is temporarily opt-in while the private project has no quota; it is not a PR gate.

Tests Added Or Updated

• tests/unit/command-code-executor.test.ts: added goPlanFallbackFetch() helper and two tests:
  • Command Code /alpha/generate fallback: reasoning-only output falls back to reasoning as content (non-stream) (#10986)
  • Command Code /alpha/generate fallback: reasoning-only output emits a content delta chunk when streaming (#10986)
• No production code changed.

Coverage Notes

• No change to src/, open-sse/, electron/, or bin/. The restored tests exercise the existing reasoning-only branches of createJsonResponse and createStreamResponse in open-sse/executors/commandCode.ts, so coverage of that file can only move up.

Reviewer Notes

• Test-only; no runtime behaviour changes and no migrations or flags.
• This does not claim to fix fix(backend): OpenAI API returns null content when model output consists of reasoning tokens (e.g. command-code/meta/muse-spark-1.2-contributor) #12130; it only re-establishes the guard that [BUG] OpenAI API returns null content when model output consists of reasoning tokens (e.g. command-code/meta/muse-spark-1.2-contributor) #10986 originally shipped so a future regression in the fallback is caught by CI.
• No changelog fragment added since there is no user-facing change; a changelog.d/maintenance/ entry can be added if preferred.