fix #375, converged flag when predefined number of iterations.

[marcocuturi]

Fix #375

New way to compute converged property of SinkhornOutput objects. It was shown that when the number of iterations was predefined (this can be done setting min_iterations=max_iterations in the solver), the output was never marked as converged (the rule to test convergence was simply whether the iteration counter had reached max_iterations.

This fix requires forces the (re?)computation of the error at the very last iteration, and to store the convergence threshold in the SinkhornOutput, to compare it when converged is called.

[codecov-commenter]

Codecov Report

Merging #386 (6c0ca9e) into main (b2b7ebb) will increase coverage by 0.65%.
The diff coverage is 100.00%.

❗ Your organization is not using the GitHub App Integration. As a result you may experience degraded service beginning May 15th. Please install the Github App Integration for your organization. Read more.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #386      +/-   ##
==========================================
+ Coverage   87.95%   88.61%   +0.65%     
==========================================
  Files          52       55       +3     
  Lines        5679     5945     +266     
  Branches      841      865      +24     
==========================================
+ Hits         4995     5268     +273     
+ Misses        561      551      -10     
- Partials      123      126       +3     

Impacted Files	Coverage Δ
src/ott/math/fixed_point_loop.py	100.00% <ø> (ø)
src/ott/tools/soft_sort.py	95.23% <ø> (ø)
src/ott/solvers/linear/sinkhorn.py	97.53% <100.00%> (+0.59%)	⬆️
src/ott/tools/sinkhorn_divergence.py	94.73% <100.00%> (ø)

... and 8 files with indirect coverage changes

[marcocuturi]

thanks for the comments @michalk8 !