optimized memmove

[ZERICO2005]

Optimized memmove, as it appears to be used a lot by C++ code.

[runer112]

LGTM. But calc84 is the true optimization specialist, so maybe give him some opportunity to re-review before accepting my approval alone.

[ZERICO2005]

Applied some optimizations suggested by calc84maniac to the src == dst optimized version. Compared with the src != dst version, the src == dst version is 1F + 1 slower when src > dst and 1F slower when src < dst.