Fix Jaro-Winkler result for two empty inputs

[mhuerster]

Resolves #43

After trying a few other libraries (https://github.com/kiyoka/fuzzy-string-match and
https://github.com/tonytonyjan/jaro_winkler), I've noticed that it seems more common to return 0
for two empty inputs than 1.

This PR implements and tests that behavior.

[Yomguithereal]

Hum. That's quite strange. Conceptually, two empty strings are identical, so the similarity between them should be 1 not 0. I don't remember but I think my library computes the Jaro-Winkler similarity, not distance. The library you cite computes the distance here and returns correctly 0 distance between two empty strings, for instance https://github.com/tonytonyjan/jaro_winkler/blob/master/test/test_jaro_winkler.rb#L42

[lenaschoenburg]

Doc strings say they compute distance, not similarity

[Yomguithereal]

@dignati you are right. There seems to be some kind of confusion in the doc strings. The function do compute the similarity, not the distance. My bad.