Added a new command "pl" in sbt to publish to both Maven and Ivy.

- SizeEstimator was following through a ClassLoader field of Hadoop
  JobConfs, which referenced the whole interpreter, Scala compiler, etc.
  Chaos ensued, giving an estimated size in the tens of gigabytes.
- Broadcast variables in local mode were only stored as MEMORY_ONLY and
  never made accessible over a server, so they fell out of the cache when
  they were deemed too large and couldn't be reloaded.

Added mapPartitionsWithSplit to the programming guide.

number of open files. Also optimized sending of disk-based blocks.

Allow controlling number of splits in distinct().

Log message which records RDD origin

custom serializers or Kryo registrators can be loaded.

This adds tracking to determine the "origin" of an RDD. Origin is defined by
the boundary between the user's code and the spark code, during an RDD's
instantiation. It is meant to help users understand where a Spark RDD is
coming from in their code.

This patch also logs origin data when stages are submitted to the scheduler.

Finally, it adds a new log message to fix an inconsitency in the way that
dependent stages (those missing parents) and independent stages (those
without) are logged during submission.

Added MapPartitionsWithSplitRDD.