Import Arm Optimized Routines v21.02

We already have copies of some of these files in the repository without a
vendor import and we should really be using contrib/ instead.
We should also be able to use some of the math functions to allow the
tests to pass on AArch64 (and other architectures) instead of just x86.
We should also be able to reuse some of the tests for the kyua testsuite.

Imported using

curl -L https://github.com/ARM-software/optimized-routines/tarball/e823e3abf5f89ecba58a10fc0fd82c13d9984b6b | tar --strip-components=1 -xvzf -
git add .