Try to apply bit-vector algorithm for computing levenshtein variants

[mrkkrp]

Here is the paper: https://pdfs.semanticscholar.org/813e/26d8920d17c2afac6bf5a15c537b067a128a.pdf. Should be faster.