Introduce cardinality limit on One Hot encodings

[klemens-floege]

Disabled by default.

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.

[gemini-code-assist]

Code Review

This pull request introduces two valuable features: capping the maximum number of features after preprocessing and adding a cardinality limit for one-hot encoding. While the intent is good, the current implementation has a few critical issues that result in the incorrect handling of feature modalities. Specifically, high-cardinality categorical features lose their categorical status after being filtered from one-hot encoding, and a new subsampling step incorrectly converts all features to numerical. The associated tests are not comprehensive enough to detect these bugs. I have provided detailed feedback and suggestions for fixes in the review comments.

[bejaeger]

Thanks!
I think the "max cardinality one-hot-encoding" is a good addition, but I think we should remove the logic for the second resampling for now.

[bejaeger]

Thanks! just one nit