NEON, AVX, SSE and C implementations for L2 distance function

input

two vectors of float have the same length.

notes:

in AVX we process 8 by 8 each shot while in SSE and NEON we process 4 by 4.
implementation of int is much faster than float and i will add it later on.