Ported a similarity calculation algorithm

[chilimangoes]

To potentially help with #49.

I ported a similarity calculation function I created to compare youtube video titles to theTVDB episode titles. For that purpose, it returned the correct episode better than 9 out of 10 times, but I don't know how well it will work for matching abbreviations. For example, if I try to match "test function", the symbol "test" has a higher score (39) than "tstfnc" (37) because of the token sequence boosting I added.

One thought I had was to change the sequence boosting to work with non-contiguous sequences. In this case, the entire sequence of "tstfnc" would count as a sequence, even though most of the letters aren't contiguous, because those letters appear in the same sequence in both.

[synkarius]

I think your proposed change to the sequence boosting is a good idea. The present PITA algorithm does something similar, also to detect non-contiguous sequences. Another thing to try might be removing all the vowels.