Add out-of-bounds detection warning to InferenceSlicer

[Copilot]

When a user's callback accidentally runs inference on the full image instead of the provided slice, supervision blindly applies slice offsets to coordinates already in full-image space — producing the same detections tiled across a grid with only N unique confidence values for N×slices total detections.

The root cause is a common callback authoring mistake:

# BUG: uses `image` (full image) instead of `image_slice`
def callback(image_slice: np.ndarray) -> sv.Detections:
    result = model.infer(image)[0]
    return sv.Detections.from_inference(result)

Changes

• InferenceSlicer._run_callback: After invoking the user callback, checks whether any returned detection coordinate exceeds the slice dimensions or is negative. Either condition indicates the callback is operating on the wrong image. Emits a SupervisionWarnings with an actionable message pointing to the likely cause.
• InferenceSlicer.__init__: Adds _out_of_slice_bounds_warned flag so the warning fires at most once per slicer instance, avoiding log spam across hundreds of slices.
• Tests: Covers warning on x/y overflow, warning on negative coordinates, single-emission across multiple slices, and no false positives for well-behaved callbacks.

---

⚡ Quickly spin up Copilot coding agent tasks from anywhere on your macOS or Windows machine with Raycast.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 77%. Comparing base (8b60ee1) to head (5a9c811).
⚠️ Report is 1 commits behind head on develop.

Additional details and impacted files

@@           Coverage Diff           @@
##           develop   #2186   +/-   ##
=======================================
+ Coverage       76%     77%   +1%     
=======================================
  Files           62      62           
  Lines         7575    7592   +17     
=======================================
+ Hits          5743    5810   +67     
+ Misses        1832    1782   -50     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

Adds a safety warning to sv.InferenceSlicer to help users detect a common callback mistake where inference is run on the full image instead of the provided slice, which would otherwise produce incorrectly offset/tiled detections.

Changes:

• Add a once-per-instance out-of-slice-bounds check after callback execution, emitting SupervisionWarnings when detection coordinates are negative or exceed slice dimensions.
• Add thread-safe “warned already” state to prevent warning spam across many slices / multiple threads.
• Add unit tests covering overflow/negative coordinates, single-emission behavior (including multithreaded), and no false positives.

Reviewed changes

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

File	Description
src/supervision/detection/tools/inference_slicer.py	Adds the bounds-check warning logic and a lock/flag to ensure it fires at most once per slicer instance.
tests/detection/tools/test_inference_slicer.py	Adds tests validating warning emission behavior and non-emission for in-bounds detections.
docs/changelog.md	Documents the new warning behavior in the changelog.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.