[FEA] RAPIDS accelerated Java UDF

[jlowe]

Is your feature request related to a problem? Please describe.
This is closely related to #1594 since they both are implemented in Spark via the ScalaUDF Catalyst class. Java UDFs however use a slightly different registration API and are implemented in terms of Spark's UDF interface. Spark wraps the object call in a lambda method which makes extracting the original user class tricky.

Describe the solution you'd like
Ideally the RAPIDS Accelerator plugin would be able to automatically extract the user class from the Java UDF wrapped in a lambda method and detect whether the user's code implements the RapidsUDF interface for executing the UDF on the GPU.