ENH: use functools.lru_cache to speed up

[Zeroto521]

Levenshtein distance algorithm can't be vectorized.
So the calculation would be very slow in large data.

An idea to speed up is using the cache.
Use the accumulate case to show the cache.

def accumulate(x):
    return sum(range(x))

accumulate(100000000) needs 5s no matter if it is the first time running or the second time running in my local without lru_cache.

from functools import lru_cache

@lru_cache
def accumulate(x):
    return sum(range(x))

After adding lru_cache, the first running accumulate(100000000) still needs 5s.
But the second time running accumulate(100000000) needs 0s.

[maxbachmann]

This is not really a good idea in terms of this library, since in most use cases it it uncommon to call the function twice with the two same strings. If this is really common in you specific case, it makes more sense to write yourself a wrapper function adding the lru_cache for your own project.