Benchmarking individual pieces of Quickenshtein

[Turnerj]

As a thought coming from my own tweet about how SSE2 being enabled didn't seem faster than no intrinsics leads me to believe that I need to better measure individual optimizations as well as everything combined.

That is to say, if I improve FillRow, I should have a FillRow benchmark. As these are private methods, I'd probably have them marked as internal and use that configuration thing to allow certain libraries access to it.

Initially, it would be for benchmarking but might even be a good idea at some really targeted testing of these components.