- Config option related to pipelining has been removed.
 - Summary: Basic -> Pipelining / Parallel -> NO pipelining

 - Renamed some ParallelLocalFileShuffle config options for clarity.

 - Renamed parallel-pull version to ParallelLocalFileShuffle
 - Note that setting max-concurrent connections to 1 in ParallelLocalFileShuffle should essentially be the same as BasicLocalFileShuffle

Enabling/disabling HTTP pipelining is a config option now. Performance tradeoffs are not obvious yet.

- Added GroupByTest.scala for benchmarking.

phase of DfsShuffle.

Fixed a small bug in DFS shuffle -- the number of reduce tasks was not being set based on numOutputSplits

well as versions of the shuffle operations that set the number of splits
automatically.

don't all hit the same nodes at the same time.

These tasks were being subjected to delay scheduling but then counted as
having been launched on a preferred node. The solution is to have a
separate queue for them and treat them as preferred during scheduling.

Conflicts:
	src/scala/spark/SparkContext.scala

Using the latest one from Matei.

are used to create an initial copy of each config file if the user does
not have one. This way, users won't accidentally commit their changes to
config files to git.

use this. Closes #12.

Simplified UnionRDD slightly and added a SparkContext.union method for efficiently union-ing a large number of RDDs

optional constructor parameter, so that the scheduler is guaranteed that
a Spark home has been set when it first builds its executor arg.

SparkContext and have the master node serve those to workers.