feat: Added support of FasterRCNN for PyTorch

[fg-mindee]

This PR introduces the following modifications:

• adds fasterrcnn_mobilenet_v3_large_fpn as supported architecture for PyTorch
• adds corresponding unittests
• updates training script accordingly (and adding AMP support)

Any feedback is welcome!

[charlesmindee]

Thanks!

[codecov]

Codecov Report

Merging #691 (73b09ee) into main (9c45ba6) will decrease coverage by 0.45%.
The diff coverage is 95.00%.

@@            Coverage Diff             @@
##             main     #691      +/-   ##
==========================================
- Coverage   96.40%   95.95%   -0.46%     
==========================================
  Files         117      120       +3     
  Lines        4540     4568      +28     
==========================================
+ Hits         4377     4383       +6     
- Misses        163      185      +22     

Flag	Coverage Δ
unittests	95.95% <95.00%> (-0.46%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
doctr/models/obj_detection/faster_rcnn/pytorch.py	93.75% <93.75%> (ø)
doctr/models/obj_detection/__init__.py	100.00% <100.00%> (ø)
doctr/models/obj_detection/faster_rcnn/__init__.py	100.00% <100.00%> (ø)
doctr/models/recognition/crnn/pytorch.py	94.18% <0.00%> (-5.82%)	⬇️
doctr/models/detection/linknet/base.py	83.16% <0.00%> (-3.97%)	⬇️
...dels/detection/differentiable_binarization/base.py	90.00% <0.00%> (-2.78%)	⬇️
doctr/models/builder.py	96.49% <0.00%> (-2.64%)	⬇️
doctr/models/recognition/master/pytorch.py	93.47% <0.00%> (-2.11%)	⬇️
...s/detection/differentiable_binarization/pytorch.py	97.50% <0.00%> (-0.81%)	⬇️
doctr/models/detection/linknet/pytorch.py	98.07% <0.00%> (+0.03%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9c45ba6...73b09ee. Read the comment docs.