Don't accept generated local directory names that can't be created

fix [SPARK-740] block manage UI throws exception when enabling Spark Streaming

RDD.unpersist: probably the most desired feature of Spark

TimeStampedHashMap.

Also updated unit tests to make sure they are properly testing for
concurrency.

Conflicts:
	core/src/main/scala/spark/SparkContext.scala

Pull request to address issues Reynold Xin reported

Conflicts:
	core/src/main/scala/spark/BlockStoreShuffleFetcher.scala

[Fix SPARK-742] Task Metrics should not employ per-record timing by default

size is 8KB in FastBufferedOutputStream, which is too small and would
cause a lot of disk seeks.

consolidate shuffle output files.

Conflicts:
	core/src/main/scala/spark/storage/DiskStore.scala

This patch does three things:

1. Makes TimedIterator a trait with two implementations (one a no-op)
2. Makes the default behavior to use the no-op implementation
3. Removes DelegateBlockFetchTracker. This is just cleanup, but it seems like
   the triat doesn't really reduce complexity in any way.

In the future we can add other implementations, e.g. ones which perform sampling.

Add zip partitions interface