[llvm] Add target-byteorder for cases where endian in target triple is what matters (PR #107915)

[Sean Perry via llvm-commits]

perry-ca wrote:

I had thought that yaml2obj would have been host independent.  It probably should be so you can generate object files for a target independent of what host you are running on.  If it is meant to be then we might need to look deeper into these tests.  The `llvm/test/Object/AArch64/filetype-fileset.yaml` does pass in all four variations (target big/little, host big/little).  I think I had missed one variation when double checking the rest.  I put the host-byteorder back in for the yaml2obj cases.

https://github.com/llvm/llvm-project/pull/107915