HttpBlockedLocalFileShuffle has also been converted to have per-reducer consumption thread. Works in local mesos, but NOT on EC2 :|

Upside: No synchronized blocking on "combiners" variable. 3x faster :)
Downside: Inefficient implementation. Requiring too much temporary data. Approx. 2x increase in memory requirement :( Should be fixed at some point.

 - Fixed config param loading bug in CustomParallelLFS

Conflicts:
	conf/java-opts
	src/scala/spark/BasicLocalFileShuffle.scala

Conflicts:
	conf/java-opts
	src/scala/spark/BlockedLocalFileShuffle.scala
	src/scala/spark/CustomBlockedLocalFileShuffle.scala
	src/scala/spark/HttpBlockedLocalFileShuffle.scala

Renamed BlockedLocalFileShuffle to HttpBlockedLocalFileShuffle for merging with the mos-shuffle branch.

 - Clients now request for byte ranges to the server using an INDEX file.

 - 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

Removed an unnecessary byte array in the middle. Probabaly will have to bring it back if we do block level data movement.

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

- Added GroupByTest.scala for benchmarking.