Lilian - sequential data toolkit Code

* Added WSJCorpusReader to read the WSJ Corpus
* Added method for generating cross-validation corpus pairs
* Added beam-search option to parser

Added new constructor to collection corpus

Touched up the matchingNodes method that checks bracket coverage, made more generic, moved it to Functions as a static method.
Fixed bug in pair that stopped equals from working with null values.

Modified the GrammarCorpus to output trees instead of tokens, so we can test the various DOP's with toy corpora. Added a Global class with a centralized RNG. removed all uses of Math.random().