[FEA] Add target_type to cuML TargetEncoder

[ZeroCool2u]

Is your feature request related to a problem? Please describe.
I would like to use the target_type='multiclass' feature of the sklearn TargetEncoder class as part of my cuML preprocessing Pipeline. This is not currently possible.

Describe the solution you'd like
I would like the target_type argument added to the cuML TargetEncoder implementation in hopes of approaching feature parity with the sklearn version of TargetEncoder.

Describe alternatives you've considered
I've tried to use the sklearn TargetEncoder implementation within a cuML Pipeline, but as expected that doesn't go well. There are some data type casting issues and I expect it would cause performance issues anyways since there would probably be some GPU <-> CPU data movement that we don't want happening there. I could also implement it myself, but that would be a good deal of work and I don't have time at the moment :(

Additional context
This is currently one of two issues preventing me from migrating my entire sklearn Pipeline over to cuML, which I very much would like to do!

[dantegd]

Thanks for the issue! I don't think this will require a lot of work, will try to prioritize this soon, though not sure when it will land exactly.