Fix various PowerPC issues

Quite a few little fixes here and there.

Even with my special handling of the copy opcode (20x speed up), the memcpy_speed test is still too slow. Mostly to do with the touch() function, which clang is smart enough to eliminate. I ended up just checking every other size. Another option would be to reduce the numbers of loops per size, or maybe not run the touch() for each iteration of function loops.

exec_parse has never been reporting any failures. So that led me to a few rules that were incorrect. Most of the rules haven't been tested too much with multiple vector elements. PowerPC right now doesn't optimize the loops to process more than one element at a time, and neither exec_opcodes_sys nor orc_test_compare_output_full test x2/x4 instructions.

I'm working on some updates for the testing and will start on adding support for multiple vector elements on PowerPC but may take a little while.

Fixes #25 (closed)