[Refactor] Rotated FCOS

[liuyanyi]

Thanks for your contribution and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily get feedback. If you do not understand some items, don't worry, just make the pull request and seek help from maintainers.

Motivation

Compared with 0.x, Rotated FCOSHead and CSLFCOSHead merged together.
Added a argument use_hbbox_loss to replace separate_angle to make it better understanding.

Before:
Arg separate_angle control the type of bbox_loss used and whether angle loss is enabled.

After:
Arg use_hbbox_loss control the type of bbox_loss. Whether angle loss is enabled is controlled by loss_angle.
In this case we can use Rotated IoU Loss and L1Loss together, Rotated IoU Loss for bbox, L1 for angle auxiliary losses.

Results:

model	Reported ap50	map	ap50	ap75
Rotated FCOS	71.28	41.10	70.87	41.21
Rotated FCOS kld	71.89	40.37	71.55	38.92
Rotated FCOS hbb csl	71.76	40.22	70.79	39.26
Rotated FCOS hbb	70.70	38.06	69.71	34.81
Rotated FCOS + l1	-	40.90	71.18	40.87

Note: The result is not stable and may have 1% ap difference. Training process may have unstable too.

We should do some exps on tricks and losses to find the best config settings.

Add l1 loss for angle, results does not appear to have improved, but it has not decreased either.

Modification

• RotatedFCOSHead
• PseudoAngleCoder
• Fix F.cosine_similarity typo in oriented_reppoints
• Fix typo of encode_size
• Fix CSLCoder empty input when keepdim=True

BC-breaking (Optional)

Does the modification introduce changes that break the back-compatibility of the downstream repos?
If so, please describe how it breaks the compatibility and how the downstream projects should modify their code to keep compatibility with this PR.

Use cases (Optional)

If this PR introduces a new feature, it is better to list some use cases here, and update the documentation.

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. The documentation has been modified accordingly, like docstring or example tutorials.

[zytx121]

LGTM

[RangiLyu]

LGTM