SPARK-1321 Use Guava's top k implementation rather than our...

SPARK-1321 Use Guava's top k implementation rather than our BoundedPriorityQueue based implementation

Also updated the documentation for top and takeOrdered.

On my simple test of sorting 100 million (Int, Int) tuples using Spark, Guava's top k implementation (in Ordering) is much faster than the BoundedPriorityQueue implementation for roughly sorted input (10 - 20X faster), and still faster for purely random input (2 - 5X).

Author: Reynold Xin <rxin@apache.org>

Closes #229 from rxin/takeOrdered and squashes the following commits:

0d11844 [Reynold Xin] Use Guava's top k implementation rather than our BoundedPriorityQueue based implementation. Also updated the documentation for top and takeOrdered.