test: add triage_calibration and file_prediction_accuracy to round-trip metric coverage in test_reaggregate.py

[decko]

Finding

File: tests/test_reaggregate.py:145
Severity: Minor
Source: Code review of PR #290

Issue

The round-trip test in test_reaggregate.py omits triage_calibration and file_prediction_accuracy from the roundtrip_metrics list. Both metrics populate sample_scores and should round-trip correctly (i.e. the mean of per-session scores should match the engine aggregate). As-is, they are silently untested.

Suggested Fix

Add "triage_calibration" and "file_prediction_accuracy" to the roundtrip_metrics list at tests/test_reaggregate.py:145. The existing if metric_name not in report.aggregate_scores: continue guard already handles the case where test samples lack data for these metrics, so no additional changes are needed.

References

• PR: feat(metrics): persist per-sample metric scores in JSON report #290 (feat: persist per-sample metric scores in JSON report)
• Related ticket: feat: persist per-sample metric scores in JSON report #259

[decko]

SODA found nothing to change — the existing round-trip test coverage is sufficient.