[llvm] [X86] Remove redundant TEST after shifts when count is non-zero (PR #169069)

[Simon Pilgrim via llvm-commits]

RKSimon wrote:

Apart from as an optsize optimisation I'm not convinced this fold is a good idea given how much trouble some CPUs have with eflags from shifts. And this is a large patch for a small gain, if it can be simplified to a small isel known-never-zero pattern for optsize I'd be a lot happier.

https://github.com/llvm/llvm-project/pull/169069