[llvm] [ARM] Prefer MUL to MULS on some implementations (PR #112540)

[Albert Huang via llvm-commits]

AlbertHuang-CPU wrote:

Thanks @davemgreen 
As checked/discussed offline, in some scenario mul is as slower as muls on M33, and STAR-MC1 is the same as M33 in terms of these behavior.
So I will add the feature also.

https://github.com/llvm/llvm-project/pull/112540