Making Rust faster than C with this one simple trick (optimization for levenshtein)

[tp971]

I have:

• Read the project README, including the benchmark descriptions

Description of changes

Replacing the inclusive ranges 1..=n and 1..=m by exclusive ranges 1..n + 1 and 1..m + 1 in the Rust version of Levenshtein improves the performance by about 25% on my machine (making it about 3% faster than the C version). The reason for that is that the implementation of inclusive ranges accommodates for the fact that the upper bound might be the largest representable integer (e.g. 1..=usize::MAX) by adding additional code that is executed in every iteration of the loop, which might also prevent certain compiler optimizations from running, making the performance even worse.

[zierf]

For me using enumerate on the string bytes instead of exclusive ranges saves ~10%.

see also PR #351