Levenshtein Automaton

[Toflar]

Implements a Levenshtein-Automaton which allows for faster trie searches by early aborting on prefixes that will never match the string. Basically, if the string is mysuperduperexample and you want to not know the distance of foobar to mysuperduperexample but just whether it matches a max distance of 2 for example, you can abort after foo because at this point, you know that foo can never match mysuperduperexample anymore.

What would be really nice to have on top:

• Different costs for insertion, replacement and deletion
• Damerau-Levenshtein support (so cost for transposition on top)
• Configurable cost for a mistake on the very first character of a string. It's common in search engines to count a typo in the first character as 2 typos.