Add ordered split to RDD class

[drJAGartner]

To perform supervised machine learning, one must partition data into labeled data sets, typically according to a set of rules. Creating an RDD function to perform this as a single step serves as a first step in moving toward single pass RDD partitioning.

[AmplabJenkins]

Can one of the admins verify this patch?

[markhamstra]

This needs to be implemented in Scala since Spark's architecture is such that the Scala implementation can be exposed via the Python, Java, R, etc. APIs and not the other way around, starting with one of those APIs as the root implementation.

[JoshRosen]

I don't think that we should add this to the core RDD API itself. This method is probably more useful / discoverable as part of one of the ml or mllib subpackages. In addition, the implementation here is not more efficient than what a user could write themselves (i.e. it's just syntactic sugar for ~ two lines of code).

If you'd still like to propose this change, please see the instructions at https://cwiki.apache.org/confluence/display/SPARK/Contributing+to+Spark and follow the process for filing a JIRA.

[drJAGartner]

Sounds good, thanks for the input.