feat(chat-api): persist review results and skip re-review for reviewed reports

[willvelida]

Summary

Updates Chat.Api's A2AReportTool to persist review results back to Reporting.Api and skip Claude re-review for already-reviewed reports. This is PR 2 of Issue #269 (Phase 2 of the ReportReviewerService fail-closed fix).

After this change, the full review persistence flow works end-to-end:

1. User asks for report status → GET /api/reports/{jobId} returns "generated"
2. Chat.Api calls Claude reviewer → persists result via PUT /api/reports/{jobId}/review
3. Next status check → GET returns "reviewed" → Chat.Api presents persisted review data without calling Claude

Changes

A2AReportTool.cs

• Split status handling: "generated" → ReviewPersistAndPresentAsync (review + PUT + present), "reviewed" → PresentPersistedReview (skip Claude, present persisted data)
• Fire-and-forget PUT: After successful Claude review, persists result via PUT /api/reports/{jobId}/review. Failure is non-blocking (logs warning, still presents review to user) per ASI08.
• New PresentPersistedReview method: Reads ReviewApproved, ReviewConcerns, ReviewValidatedSummary from GET metadata. Approved reports show validated summary; flagged reports show concerns list.
• Updated ReportMetadataDto: Added 4 review fields (ReviewApproved, ReviewConcerns, ReviewValidatedSummary, ReviewedAt) to deserialize persisted review data.

Tests (6 new)

• PersistReviewResult_WhenStatusIsGenerated_AndReviewCompleted — verifies PUT called
• SkipPersistence_WhenStatusIsGenerated_AndReviewNotCompleted — verifies PUT NOT called
• PresentPersistedReview_WhenStatusIsReviewed_AndApproved — verifies reviewer NOT called
• PresentPersistedConcerns_WhenStatusIsReviewed_AndNotApproved — verifies concerns in output
• HandlePutFailureGracefully_WhenPersistenceFails — PUT throws, result still returned
• HandlePut409Gracefully_WhenAlreadyReviewed — PUT returns 409, result still returned

Validation

• 188 tests pass (170 unit + 15 integration + 3 evaluation), 0 failures
• 82.9% line coverage (threshold: 70%)
• GetReportStatusTool (obsolete) was NOT modified — only A2AReportTool was updated

Context

• Issue: Research: Persist review results to blob storage + wire up ReportStatus.Reviewed #269
• Depends on: PR feat(reporting-api): add PUT review endpoint and wire ReportStatus.Reviewed #303 (Reporting.Api PUT endpoint) — already merged
• Key design decision: PUT is fire-and-forget with try/catch. If persistence fails, the user still sees the review (graceful degradation, ASI08). The in-memory IMemoryCache from Phase 1 acts as defense-in-depth.

[Copilot]

Pull request overview

Updates Biotrackr.Chat.Api’s A2AReportTool to persist Claude review results back to Reporting.Api and to skip re-review when a report is already in the "reviewed" state, presenting persisted review metadata instead.

Changes:

• Split "generated" vs "reviewed" status handling to either (a) review+persist+present or (b) present persisted review directly.
• Added best-effort PUT /api/reports/{jobId}/review persistence after a completed review.
• Extended ReportMetadataDto and added unit tests for persistence + skip behavior.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
src/Biotrackr.Chat.Api/Biotrackr.Chat.Api/Tools/A2AReportTool.cs	Adds review persistence PUT and introduces a fast-path to present persisted review data for "reviewed" reports.
src/Biotrackr.Chat.Api/Biotrackr.Chat.Api.UnitTests/Tools/A2AReportToolShould.cs	Adds unit tests covering PUT persistence, skip persistence, skip reviewer on reviewed status, and graceful handling of PUT failures/409.

[github-actions]

Package	Line Rate	Branch Rate	Health
Biotrackr.Chat.Api	83%	65%	✔
Summary	83% (763 / 923)	65% (127 / 196)	✔

Minimum allowed line rate is 70%