give segmentation QA the same threshold-independent view as detection

Detection QA reports a precision/recall curve, average precision and a
calibration sweep; segmentation QA reported a single operating point. Both rank
their outputs by confidence, so the same view applies, and the asymmetry meant
the two panels answered different questions about comparable runs — an
inconsistency introduced when detection gained the curve.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

docs/API_CONTRACTS.md

@@ -1981,7 +1981,12 @@ compare two models whose calibration differs; the curve can. `average_precision`
 existing ones.
 
 Segmentation QA (`POST /api/v1/segmentation/runs/{analysis_run_id}/qa/reference`)
-applies the same tile-coverage clipping and returns the same `coverage` block.
+applies the same tile-coverage clipping and returns the same `coverage` block,
+and now also the same `precision_recall_curve`, `calibration_thresholds` /
+`calibration_sweep` and average-precision metrics. Segmentation confidences rank
+the same way detection confidences do, so the same threshold-independent view
+applies; without it the two QA panels answered different questions about
+comparable runs.
 Without it, every reference feature outside the inferred tiles counted as a
 false negative and recall was understated by an arbitrary amount.