AVX critical for good performance on modern CPUs

[Jack Hale]

At the moment a user taking HPGMG-FE and compiling per the instructions on the README might get a sub-optimal binary because the AVX fused multiply add routines are not enabled. The vast majority of in-use modern x86-64 CPUs have AVX available.

Suggestions: A note in the README that the CFLAGS are likely to be suboptimal. Perhaps an additional file where people can record some good starting flags for common toolchains.