[RFC] Scikit interface for the `ml` and `mllib` packages

[kszucs]

PySpark machine learning packages are getting more robust. It is considerable to make use of the already implemented distributed algorithms through an sklearn compatible interface instead of porting the non-distributed ones.