With the current approach that cmake makes uses for universal binaries
getProcessTriple was returning the same value whether or not being run on
x86_64 or arm. This resulted in failures if one ran the tools/tests with
arch-x86_64 or arch -arm64 (depending on LLVM_HOST_TRIPLE that was set at build
time). Change getProcessTriple to consider the architecture being compiled when
populating triple.
I'm not sure if this is best approach but LLVM_HOST_TRIPLE is fixed per
compilation given current universal binary approach and this is noninvasive
while enabling usage inside universal binary so easy to change.
Is the test useful here?