Add optimizations for the AMD genoa and bergamo chips

These have AVX512. With the Intel compilers you need to be careful not to use -x optimizations so it isn't straight-forward to get these optimizations, faking skylake-avx512 seems to work.

Note that using -axCORE-AVX512 alone does not work, that gives a binary with two code paths, a baseline version and the optimized one. The optimized one is only selected when running on Intel CPUs, so we need this extra suggestion to make that work on AMD.