[Refactor] PyTorch SAR_Resnet31 make it ONNX exportable (again)

[felixdittrich92]

This PR:

• reverts pooling to fix onnx export (retrained with same results as before)
• set attention units to 512 as described in paper (fix my mistake ^^)
• some formatting / docu

Any feedback is welcome 🤗

(functional pooling has raised some ONNX export issues)

[codecov]

Codecov Report

Merging #930 (d92ba7b) into main (0c8dd60) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #930      +/-   ##
==========================================
+ Coverage   94.68%   94.69%   +0.01%     
==========================================
  Files         134      134              
  Lines        5491     5490       -1     
==========================================
  Hits         5199     5199              
+ Misses        292      291       -1     

Flag	Coverage Δ
unittests	94.69% <100.00%> (+0.01%)	⬆️

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

Impacted Files	Coverage Δ
doctr/models/recognition/sar/pytorch.py	98.50% <100.00%> (-0.02%)	⬇️
doctr/transforms/functional/base.py	97.10% <0.00%> (+1.44%)	⬆️

---

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 0c8dd60...d92ba7b. Read the comment docs.

[frgfm]

Thanks Felix! Just a small comment perhaps?

[frgfm]

Looks good, thanks!