CentripetalNet Support

[dnth]

This PR adds CentripetalNet with HourglassNet backbone to IceVision. Model only works for object detection task. No mask related config found on the mmdet repo for instance segmentation task.
The following screenshot shows COCOMetric during training using the getting started notebook on my local machine.

Also sharing the training on Colab
https://colab.research.google.com/drive/1ADQ8mts1GRc-_y1CqGMvWfZ7ipK4LfWV?usp=sharing

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

Merging #1050 (67ad629) into master (eb550f8) will increase coverage by 0.01%.
The diff coverage is 80.95%.

@@            Coverage Diff             @@
##           master    #1050      +/-   ##
==========================================
+ Coverage   85.92%   85.94%   +0.01%     
==========================================
  Files         277      280       +3     
  Lines        5835     5934      +99     
==========================================
+ Hits         5014     5100      +86     
- Misses        821      834      +13     

Flag	Coverage Δ
unittests	85.94% <80.95%> (+0.01%)	⬆️

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

Impacted Files	Coverage Δ
icevision/models/mmdet/utils.py	56.71% <20.00%> (-1.10%)	⬇️
...ion/models/mmdet/models/centripetalnet/__init__.py	100.00% <100.00%> (ø)
.../mmdet/models/centripetalnet/backbones/__init__.py	100.00% <100.00%> (ø)
...mmdet/models/centripetalnet/backbones/hourglass.py	100.00% <100.00%> (ø)
icevision/data/__init__.py	100.00% <0.00%> (ø)
icevision/data/convert_records_to_fo.py	85.89% <0.00%> (ø)
icevision/soft_dependencies.py	96.29% <0.00%> (+0.14%)	⬆️

---

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 eb550f8...67ad629. Read the comment docs.