[AArch64] Add optimized strchr.

Implementation of strchr for AArch64.  Speedups taken from micro-bench
show the improvements relative to the standard C code.

The use of LD1 means we have identical code for both big- and
little-endian systems.