feature/v2

[du-phan]

This PR adds new components, several enhancements and bug fixes.

New components:

• A recipe to compute drift between a deployed model and a dataset in the flow.
• A recipe to compute drift between two datasets in the flow.
• A custom metric that retrieve the most recent drift score in the resulted dataset from the two recipes above.

Main enhancements:

• Add support for regressions in the model view.
• Hide unnecessary information on model view when needed.
• Add support for all algorithms: for those that are not tree-based (and thus can not be used in v1), we use a surrogate model to get the feature importance information.
• Add riskiest features information allowing user to immediately have the list of features that they need to be careful about (ie. features that are drifted the most and are important in the deloyed model).
• Add support for partitioning.
• Add support for all types of train-test split (with/without cross-validation).

Main bug fixes:

• Fix bug with boolean dtype handling that leads to mismatch prediction probability and weird categorical variable encoding.
• Fix bug with Date dtype and python 3.

[du-phan]

@MayeulRousselet I've addressed all your comments in the recent commits. Back to you thank you 🙏

[du-phan]

@MayeulRousselet I finally managed to finish this round, back to you 🙏