[Perf] Update inlining memcpy patch (#563)

The patch file for setting inilining memcpy preference was slightly
incorrect. Use of the SmallVector data structure caused issues with
correctly inlining on some tests. The optimisation was also not enabled
for v8.1m mainline targets as it should be. I've updated the patch file
to reflect this.

Author: nasherm

Omax.cfg

@@ -8,3 +8,4 @@
 -mllvm -enable-dfa-jump-thread \
 -mllvm -enable-loop-flatten \
 -mllvm -enable-unroll-and-jam \
+-mllvm -enable-inline-memcpy-ld-st