Add support for mmlspark

It would be a useful package to support as it's currently the only spark based package that supports class weights for tree based algorithms.

Start by reviewing existing Spark package integrations in the mleap repository and identify the entry points used to add package support. Clarify the required mmlspark scope, compatibility expectations, and validation approach with maintainers. Done should mean mmlspark is supported and its class-weighted tree-based algorithms are covered by appropriate validation.

Written by the indexing model from the issue text.