fix: ID=0 bug in Mean Average Precision calculation

[balthazur]

Description

Objects were getting 0.0 mAP despite perfect IoU matches due to a bug in annotation ID assignment. The COCO evaluator treats ID=0 as "no match", but supervision was assigning the first annotation ID=0, causing it to be incorrectly classified as a false positive.

Root Cause

# BUGGY: Started IDs from 0
"id": len(annotations),  # First object gets ID=0 → treated as "no match"

Solution

# FIXED: Start IDs from 1
"id": len(annotations) + 1,  # Start IDs from 1 (0 means no match)

Impact

• Before: Single objects → 0.000 mAP, Multi-objects → partial failure
• After: All perfect predictions → 1.000 mAP

Testing

Added 3 regression tests covering single objects, multiple objects, and batch updates to prevent future regressions.

Reproduction Notebooks:

supervision 0.26.0 showing the bug:
https://colab.research.google.com/drive/1naVTb7qo7rbgT_AYaD8V9CvRdhK_wwPH?usp=sharing

With fix:
https://colab.research.google.com/drive/1Kz34ec1zuDbqE7lovvTWz1REKLNsyPuM?usp=sharing

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?

Tests, and in a reproduction notebook

Any specific deployment considerations

For example, documentation changes, usability, usage/costs, secrets, etc.

Docs

• Docs updated? What were the changes: