fix(eval): preserve custom eval metadata

[he-yufeng]

Summary

• allow eval models to carry custom metadata fields instead of rejecting them as extras
• cover both top-level EvalCase metadata and nested SessionInput metadata in serialization

Fixes #5906

To verify

• PYTHONUTF8=1 uv run --with pytest pytest tests/unittests/evaluation/test_eval_case.py -q
• PYTHONUTF8=1 uv run ruff check src/google/adk/evaluation/common.py tests/unittests/evaluation/test_eval_case.py
• PYTHONUTF8=1 uv run pyright src/google/adk/evaluation/common.py
• PYTHONUTF8=1 uv run python -m py_compile src/google/adk/evaluation/common.py tests/unittests/evaluation/test_eval_case.py
• git diff --check

[adk-bot]

I have completed the analysis of Pull Request #5908 ("fix(eval): preserve custom eval metadata"), which resolves Issue #5906.

I verified the contributor's Google CLA signature and performed an architectural/style review against the current repository state:

• Checked EvalBaseModel baseline which was indeed configured with extra="forbid".
• Reviewed the PR's implementation of changing this configured value to extra="allow".
• Assessed the newly added test test_eval_models_preserve_extra_metadata, finding that it adheres to ADK's testing standards.

Please review the comprehensive findings and detailed checklist in our generated report:

👉 pr_analysis_report.md

Key Decisions / Actions For You:

• No blockers identified: The change is perfectly aligned, safe, and backwards-compatible.
• Next Steps: You may approve and merge this pull request in the repository. Let me know if you would like

[rohityan]

/adk-pr-analyze

[adk-bot]

I have completed the read-only triage and architectural review of Pull Request #5908 ("fix(eval): preserve custom eval metadata"), which resolves Issue #5906.

Please review the comprehensive findings and detailed checklist in our generated report:

👉 pr_analysis_report.md

Key Highlights & Decisions:

• No architectural blockers found: The transition of EvalBaseModel's model_config from extra="forbid" to extra="allow" matches target Pydantic patterns perfectly and retains backwards compatibility.
• Recommended Action: Approve (or Approve with Nits to request adding a behavioral docstring to the new test test_eval_models_preserve_extra_metadata in compliance with Rule 2 of the ADK Testing Style Guide).

[rohityan]

Hi @he-yufeng , Thank you for your contribution! We appreciate you taking the time to submit this pull request.

[he-yufeng]

Closing this older branch in favor of the cleaned replacement PR #5922 for the same issue and same fix. #5922 has been rebased onto current main and keeps the diff limited to the intended two files.