Preprocessing - Consider Catboost for target encoding

[JanBenisek]

Insipire by https://catboost.ai/ for target encoding. To be investigated

[MatthiasRoels]

The implementation used in Catboost is tightly coupled with building the decision trees itself. So decoupling them will be an issue. But, you could draw inspiration from the way the encoded values are calculated to provide an alternative to the current implementation.