Faster-coco-eval

- [x] Support CrowdPose Dataset
- [x] add CrowdPose simple test
- [x] ImageEvaluation pickleable https://github.com/MiXaiLL76/faster_coco_eval/issues/45
- [x] Create test for MP eval https://github.com/MiXaiLL76/faster_coco_eval/blob/main/tests/test_world_evalutor.py

- [x] fix windows mask api.
- [x] numpy>=2 support for py39+
- [x] update CI\\CD for build macos & windows whl

- [x] fix windows & macos build for [torchmetrics](https://github.com/Lightning-AI/torchmetrics/pull/2750)

- [x] Rework [mask_api](csrc/mask_api/mask_api.cpp) with pybind11 C++ .
- [x] Rework [RLE support](csrc/mask_api/src/rle.cpp).
- [x] Create test files for all components.
- [x] The math_matches function has been reworked, with an emphasis on using C++ code.
- [x] Added more documentation of functions. Optimized existing ones.
- [x] Added rleToBoundary func with 2 backend \["mask_api", "opencv"\]
- [x] IoU type [boundary](https://github.com/bowenc0221/boundary-iou-api/tree/master) support (further testing is needed)
- [x] Create async rle and boundary comput [discussion](https://github.com/MiXaiLL76/faster_coco_eval/pull/31#issuecomment-2308369319)

- [x] Compare COCOEval bug fix.

- [x] Replace CED MSE curve with MAE (px) curve
- [x] Add CED examples
- [x] Display IoU and MAE for keypoints
- [x] Reworked eval.\_prepare to clear up the return flow
- [x] Reworked the C++ part of **COCOevalEvaluateImages** and **COCOevalAccumulate**
- [x] Add new **COCOevalEvaluateAccumulate** to combine these two calls. You can use old style **separate_eval==True** (default=False)
- [x] **COCOevalAccumulate** & **COCOevalEvaluateAccumulate** -> *COCOeval_faster.eval* is now correctly created as numpy arrays.
- [x] Append LVIS dataset support **lvis_style=True** in COCOeval_faster

py
cocoEval = COCOeval_faster(cocoGt, cocoDt, iouType, lvis_style=True, print_function=print)
cocoEval.params.maxDets = [300]