[MRG] Make gromov loss differentiable wrt matrices and weights

[rflamary]

Types of changes

• Docs change / refactoring / dependency upgrade
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Motivation and context / Related issue

Differentiable gromov will make it easier to use with gradient descent algorithm with the backends.

How has this been tested (if it applies)

All existing unit tests + new testing tat gradients are computed

Checklist

• The documentation is up-to-date with the changes I made.
• I have read the CONTRIBUTING document.
• All tests passed, and additional code has been covered with new tests.

[codecov]

Codecov Report

Merging #302 (a0698d4) into master (9c6ac88) will increase coverage by 0.04%.
The diff coverage is 94.11%.

@@            Coverage Diff             @@
##           master     #302      +/-   ##
==========================================
+ Coverage   93.37%   93.41%   +0.04%     
==========================================
  Files          21       21              
  Lines        4799     4848      +49     
==========================================
+ Hits         4481     4529      +48     
- Misses        318      319       +1