Ballista: Implement map-side of shuffle

[andygrove]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.
As a small step towards implementing #63 it would be good to implement a feature where a query stage can apply hash partitioning to the data produced by its plan and write one IPC file per output partition. This implements the map-side of the shuffle operation.

Describe the solution you'd like
QueryStageExec should have an optional output partitioning scheme, restricted to accepting hash partitioning only. If this is specified then the batches produced by the query stage's plan would be hash-partitioned and written to separate files on disk.

Describe alternatives you've considered
None

Additional context
None