Applies the same fixes to multiply blending in the AVX2
optimized path that were previously applied to the SSE2 path.
Namely, alpha blending and colors now both blend correctly
The way the SIMD blitting paths were previously organized
ensured that the AVX2 one would never actually build. This
commit separates the different implementations into
classes with differing names, ensuring the shadowing
cannot occur.
Kaloyan Chehlarski
Wyatt Radkiewicz