Ai-privacy-toolkit

Two methods for assessing privacy of synthetic datasets have been added: DatasetAttackMembershipKnnProbabilities that is based on distances of members (training set) and non-members (holdout set) from their nearest neighbors in the synthetic dataset, and DatasetAttackWholeDatasetKnnDistance that measures the share of synthetic records closer to the training than the holdout dataset.

Generic wrappers for datasets and models to enable framework independence of code.
Anonymization and minimization assets updated to use these wrappers (breaking changes to some APIs).

Anonymization + minimization modules supporting categorical features and regression models.