Faster dorogovtsev mendes

[AntoineBut]

Addresses #370.

This implementation has $O(N)$ runtime, while the previous implementation had $O(N^2)$ runtime.

The optimization comes from storing all edges added at each iteration in an indexed table, allowing to draw a random edge in $O(1)$ time, while the previous implementation had to go through (on average) half on the nodes in order to draw a random edge, which takes $O(N)$ time.

On my computer, this implementation uses approximately 16% more memory.

It runs faster than the previous one even on small graphs ($N = 100$).

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.28%. Comparing base (afb8245) to head (b42ab1f).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #371   +/-   ##
=======================================
  Coverage   97.28%   97.28%           
=======================================
  Files         118      118           
  Lines        6876     6876           
=======================================
  Hits         6689     6689           
  Misses        187      187           

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

[gdalle]

Failures on nightly are expected

[gdalle]

Thanks for the fix, this is nearly good to go!