Adds NearestDAWG for Levenshtein distances

[poke1024]

Based on http://stevehanov.ca/blog/?id=114

The relevant commit here is:
7a2379a

Sort of depends on #38.

Allows you to compute fast Levenshtein distances over large vocabularies:

import dawg

keys = ["weather", "this", "there", "thus", "another"]

d = dawg.NearestDAWG(keys)

for k, cost in d.items("other", 2):
	print(k, cost)

With #38, this could be a single new function similar inside the DAWG class (my preferred solution).

[poke1024]

This needs more work. I'll continue to work on this on a fork.