Fuzzy intersection using Levenshtein alignment

[chrislit]

Add another intersection type: Fuzzy intersection based on ordered tokens, using Levenshtein alignments to parcel out the similarity weights: Given two aligned strings:

• two equal tokens mean weight for the token is divided by 2 and added to the intersection
• two unequal tokens mean weight for each token is added to its non-intersection, except in the cases of '-' tokens which don't accrue weight since they represent ins/del