[Feature] Support OC Cost metric

[okotaku]

Motivation

github: https://github.com/mayu-ot/oc-cost paper: https://arxiv.org/abs/2203.14438

Use cases (Optional)

val_evaluator = dict(
    type="CocoOCCostMetric",
    alpha=0.5,
    beta=0.6,
    ann_file=ann_file,
    metric="bbox",
)

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
2. The modification is covered by complete unit tests. If not, please add more unit test to ensure the correctness.
3. If the modification has potential influence on downstream projects, this PR should be tested with downstream projects, like MMDet or MMPreTrain.
4. The documentation has been modified accordingly, like docstring or example tutorials.

[hhaAndroid]

@okotaku Thank you. You can create a folder named "mmdetection/configs/occost" and provide evaluation configurations for several classical algorithms(faster rcnn/retinanet/tood/dino) in it. Additionally, you can also provide a configuration for joint evaluation with COCO mAP. Afterwards, you can include the specific metrics in the README for future reference.

[okotaku]

@hhaAndroid This PR is ready!

[hhaAndroid]

@okotaku I'm sorry for the delayed response. Now that there is a conflict, please synchronize and update me. Thank you.

[okotaku]

@hhaAndroid This PR is ready.