fix: correct area handling in MeanAveragePrecision for size-specific mAP

[balthazur]

Description

This PR fixes an error in supervision.metrics.mean_average_precision.MeanAveragePrecision where the area used for size-specific evaluation (small / medium / large) was always zero unless explicitly provided in Detections.data.

The fix now:

1. Checks data["area"] if present (COCO import path).
2. Otherwise falls back to Detections.area[target_idx], leveraging Supervision’s property which already handles both bbox and mask cases.

In addition, a lightweight test suite (test/metrics/test_mean_average_precision_area.py) has been added to guard against regressions.

To reproduce:

Use a non-coco dataset as ground-truth (in my case yolov5). I am using from_yolo followed by using the Detection objects as GT's. Than, create the predictions using from_inference with results from our API. Pass both to the mAP calculator.

Before:

The areas will be 0 (the default fallback), leading to wrong/non-existing metrics for the specific sizes.

After:

Dependencies: none.

Type of change

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• This change requires a documentation update

How has this change been tested, please provide a testcase or example of how you tested the change?

Running tests and testing with a yolov5 dataset.

The new tests cover:
Normal detections (empty data) – area is auto-calculated.
COCO detections (area in data) – provided area is preserved.
Mixed batch – both code paths together.
Size-specific mAP no longer returns -1.0 for medium / large.
Verification that the evaluator truly uses Detections.area.
All existing tests continue to pass.

Any specific deployment considerations

None – pure Python change, no API or dependency impact.

Docs

• Docs updated? What were the changes: