Add Weissman score calculation

[chrislit]

W = alpha * r/rb * log(Tb)/log(T)

r/T are compression ratio & compression time for the target algorithm
rb/Tb are compression ration & compression time for a standard algorithm (in this case use gzip)

Develop a similar metric for phonetic algorithms--
involving T (obviously) and something like vocabulary(?)
standard algorithm == American Soundex?

Develop a similar metric for distance--
involving T (obviously) and ... something like normalized distance(?)
standard algorithm == normalized Levenshtein

[chrislit]

This will go in narmer.stats:
chrislit/narmer#1