feat: rtree_rotated_nms

[gcanat]

Ready for review I think.

• added Rtree nms for rotated boxes
• added a fast path to rotated_nms: check if envelopes interesect before computing expensive rotated intersection (2x speed-up it seems)
• factorized the filtering and sorting of boxes into filter_and_sort_scores function. Took the liberty of removing the n argument from the axis-aligned nms functions (hope it's ok)

[codspeed-hq]

Merging this PR will improve performance by ×2.8

⚡ 21 improved benchmarks
✅ 146 untouched benchmarks
🆕 8 new benchmarks
⏩ 1 skipped benchmark¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	test_rotated_nms[float64]	278.1 µs	151.8 µs	+83.2%
⚡	test_rotated_nms[int16]	285.3 µs	153.1 µs	+86.41%
⚡	test_rotated_nms[float32]	291.1 µs	151.3 µs	+92.4%
⚡	test_rotated_nms[int64]	288.2 µs	151.4 µs	+90.38%
⚡	test_rotated_nms_many_boxes[10000]	2.8 s	1.2 s	×2.3
⚡	test_rotated_nms[int32]	284.9 µs	150.1 µs	+89.83%
⚡	test_rotated_nms[uint64]	372.7 µs	226.7 µs	+64.37%
⚡	test_rotated_nms[uint32]	309.7 µs	158.6 µs	+95.22%
⚡	test_rotated_nms[uint8]	314.7 µs	182.5 µs	+72.39%
⚡	test_rotated_nms[uint16]	304.1 µs	156.2 µs	+94.71%
⚡	test_rtree_nms[int32]	144.4 µs	88.6 µs	+62.98%
⚡	test_rtree_nms[int16]	129.4 µs	82.9 µs	+56.22%
⚡	test_rtree_nms_many_boxes[10000]	20.5 ms	7.6 ms	×2.7
⚡	test_rtree_nms[int64]	148.4 µs	92.1 µs	+61.17%
⚡	test_rotated_nms_many_boxes[5000]	691.8 ms	291.2 ms	×2.4
⚡	test_rotated_nms_many_boxes[1000]	28.9 ms	12.3 ms	×2.4
⚡	test_rtree_nms[float64]	151.2 µs	91.6 µs	+64.99%
⚡	test_rtree_nms[float32]	153.6 µs	90.2 µs	+70.42%
🆕	test_rtree_rotated_nms[int32]	N/A	119.9 µs	N/A
⚡	test_rtree_nms_many_boxes[20000]	44.1 ms	15.9 ms	×2.8
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

---

_{Comparing gcanat:rtree_rotated_nms (350cf98) with main (5e3b449)}

Footnotes

1. 1 benchmark was skipped, so the baseline result was used instead. If it was deleted from the codebase, click here and archive it to remove it from the performance reports. ↩

[codecov]

Codecov Report

❌ Patch coverage is 99.04762% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 95.00%. Comparing base (db09f08) to head (350cf98).
⚠️ Report is 24 commits behind head on main.

Files with missing lines	Patch %	Lines
powerboxesrs/src/nms.rs	98.58%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #76      +/-   ##
==========================================
- Coverage   98.77%   95.00%   -3.78%     
==========================================
  Files          17       13       -4     
  Lines        2703     2745      +42     
==========================================
- Hits         2670     2608      -62     
- Misses         33      137     +104     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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

[Smirkey]

lgtm, will have a deeper look this we
i'll probably add a test against torch nms to verify we're correct,
been doing this manually, it's not great