[Refactor] refactor RetinaNet based on boxlist

[zytx121]

Review sequence：

1. [Refactor] Refactor anchor head and base head with boxlist mmdetection#8625 (@BIGWangYuDong @RangiLyu)
2. [Refactor] refactor RetinaNet based on boxlist #472 (@BIGWangYuDong @hhaAndroid @RangiLyu)
3. [Refactor] Refactor standard roi head with boxlist. mmdetection#8658 (@BIGWangYuDong @Czm369 @RangiLyu)
4. [Refactor] refactor Rotated Faster RCNN #487 (@BIGWangYuDong @Czm369 @RangiLyu)
5. [Refactor] Rotated ATSS with centerness loss #496 (WIP)
6. [Refactor] refactor Rotated RepPoints #497 (WIP)

• Align test
• Align train
• UT

Reviewer

Experiments

rotated_retinanet_hbb_r50_fpn_1x_dota_oc.py
report AP50: 64.55
---1---
AP50: 0.6536518893464965
AP75: 0.37845625864439075
mAP: 0.3748858855015227
---2---
AP50: 0.6475838892089881
AP75: 0.36369350507433457
mAP: 0.36750970093443525
---3---angle of anchor = 0
AP50: 0.6501743535834839
AP75: 0.3691745641139588
mAP: 0.3720109012496605
---4---angle of anchor = 0
AP50: 0.6475716856907938
AP75: 0.38000729085075463
mAP: 0.37155904887837554

rotated_retinanet_obb_r50_fpn_1x_dota_le90.py
report AP50: 68.42
---1---
AP50: 0.6813613319198807
AP75: 0.40674754402438273
mAP: 0.39424048001422823
---2---
AP50: 0.6798798602380868
AP75: 0.4070814841151892
mAP: 0.3940447051092699
---3--- angle of anchor = 0
AP50: 0.6908426650447911
AP75: 0.431021863367571
mAP: 0.4129368419550853
---4--- angle of anchor = 0
AP50: 0.688470208702759
AP75: 0.4245977440716145
mAP: 0.4105707512242186

rotated_retinanet_obb_r50_fpn_1x_dota_le135.py
report AP50: 69.79
---1---
AP50: 0.6824654196014557
AP75: 0.3393284531284533
mAP: 0.3748612843648386
---2---
AP50: 0.6782950228188119
AP75: 0.338945173004708
mAP: 0.3783645578481173
---3--- angle of anchor = 0
AP50: 0.6916534816759408
AP75: 0.36711282194397765
mAP: 0.39332846553326106
---4--- angle of anchor = 0
AP50: 0.6916534816759408
AP75: 0.36711282194397765
mAP: 0.39332846553326106