Import an optimized arm64 memcmp into the kernel

Bring in a version of the Arm Optimized Routines memcpy from before
the VFP registers were used.

Imported with modification from:
https://github.com/ARM-software/optimized-routines/blob/e823e3abf5f89ecb/string/aarch64/memcmp.S

Sponsored by: The FreeBSD Foundation