feat(datadrift): Adding data drift module

[vcerqueira]

Adds a data drift module next to the existing concept drift detectors.

Concept drift detectors watch a scalar error signal (REQUIRES_FIT = False). Data drift detectors compare incoming features to a reference distribution (REQUIRES_FIT = True). Both families stay in the capymoa.drift.detectors namespace.

The shared base class handles the sliding test window, per-feature tests, Bonferroni correction, and detection bookkeeping. A reference is set with fit(X_ref) or collected from the stream with auto_fit_samples.

14 detectors:

• Statistical tests: Kolmogorov–Smirnov, Anderson–Darling, Cramér–von Mises, Chi-square
• Distance-based: KL, Jensen–Shannon, PSI, Hellinger, Wasserstein, energy distance
• Multivariate: MMD, D3, IBDD
• Bayesian (per feature): BNDM

Also adds DataDriftResult (statistic, optional p-value/distance, per-feature flags), a tutorial notebook (notebooks/data_drift_detection.ipynb), and unit tests for instantiation, fit, detection, and get_params().

Also extends tests/test_detector.py to cover basic aspects of data drift methods.

[tachyonicClock]

Hi @vcerqueira. Great work on this. New methods yay 🤤 .

I have fixed the doc issues for you, since I know how painful they are when your not familiar with sphinx. The change was adding mapping to external docs scipy, and the :meta public: tag to ensure the "protected" method _fit/_test shows up in the docs. By default "private" methods are hidden creating issues when you want a "protected" style method.

I took some screen shots for @vcerqueira / @hmgomes to review. Are you happy with where everything is. I think you did a great job with the documentation. I'm confident a user will be able to figure out how to wire everything up.
I will, with @vcerqueira permission, move some things around in the future to achieve my standardization and organizational ambitions!

[vcerqueira]

Hi Anton @tachyonicClock ,

Thanks!! The images look good. And of course, move things around as you see fit.
Thanks for the review, cheers.

[hmgomes]

Hi @vcerqueira , can you check the backlog? I have just put out a more comprehensive review there, but regardless this is looking good already

[vcerqueira]

Sure, thanks. Will take care of it soon