Implement better BKTree distance #38
Description
Typically we would need proper Damerau–Levenshtein distance, or better a normalized one! But these algorithms are cruel they work with -1 based arrays so we need to create our own data type first that can shift base index in standard vectors delegate all vec functionality.