Optimize star coloring and its postprocessing

[amontoison]

• Preallocate the number of entries (i.e., the number of edges) in the dictionary star.
• Add a vector of integers nb_spokes to count the number of spokes for each star during the star coloring.
• Use nb_spokes to know the size of each vector in spokes when creating a StarSet.
• Restore the negative hubs in the trivial stars so that we can treat them after the "normal stars" during postprocessing.
• Improve handling of trivial stars during postprocessing by checking if a vertex is already a hub in a different star.
• Improve _update_stars! to minimize duplicated access in the dictionary star.

[amontoison]

@gdalle The failures are related to the decompression of bicoloring.
It should be related to the neutral color and the function postprocess! that I optimized.
I think that we neutralized some colors now with the random tests.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00%. Comparing base (7149c41) to head (d8bcb71).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main      #171   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           13        13           
  Lines         1423      1447   +24     
=========================================
+ Hits          1423      1447   +24     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[amontoison]

The gain in terms of storage is between 24% and 67% for the star coloring.