[SPARK-14831][SPARKR] Make the SparkR MLlib API more consistent with Spark

## What changes were proposed in this pull request?

This PR splits the MLlib algorithms into two flavors:
 - the R flavor, which tries to mimic the existing R API for these algorithms (and works as an S4 specialization for Spark dataframes)
 - the Spark flavor, which follows the same API and naming conventions as the rest of the MLlib algorithms in the other languages

In practice, the former calls the latter.

## How was this patch tested?

The tests for the various algorithms were adapted to be run against both interfaces.

Author: Timothy Hunter <timhunter@databricks.com>

Closes #12789 from thunterdb/14831.