Implementation of the incremental Kolmogorov-Smirnov Statistics

[hoanganhngo610]

This implementation features the incremental KS statistics, which is first proposed in the KDD 2016 paper.

This implementation is based on Treap (or Cartesian Tree), with bulk operation and lazy propagation. This allows the implementation to conduct insertion and removal operations in $O(log N)$ with high probability and calculate the KS test in $O(1)$, a significant improvement compared to the $O(N log N)$ within the non-incremental implementation.

As said, this implementation supports both insertion (via update) and deletion (via revert). Moreover, a closely related version of the KS Statistics, the Kuiper test, can also be calculated by setting statistic to kuiper.

[hoanganhngo610]

@MaxHalford Would you mind having a look through the PR? I think the only remaining problem would be the fact that ruff is continuously failing. (Update: Issue resolved by bumping ruff version locally from 0.1.3 to 0.1.2).

[MaxHalford]

Great job @hoanganhngo610! It's great that you added a unit test, it makes me trust the code is correct.

Can you add an entry to UNRELEASED.md?

[hoanganhngo610]

@MaxHalford I have successfully incorporated all of your proposed changes to the PR, and also provided a description within the UNRELEASED.md file. Would you mind if I merge the PR into the repository after all tests have passed?

[MaxHalford]

Merged @hoanganhngo610, good job! It's really nice to have this :)

[hoanganhngo610]

Thank you so much @MaxHalford! That's really encouraging to hear!