[llvm] [AMDGPU] Make rotr illegal (PR #166558)

[Aleksandar Spasojevic via llvm-commits]

aleksandar-amd wrote:

> @aleksandar-amd what do you think about this as an alternative to #143551?
Taking into account that https://github.com/llvm/llvm-project/pull/165295 resolved optimal selection of fshr with uniform operands, making rotr illegal and using fshr, which is a more powerful instruction, seems fine to me.

https://github.com/llvm/llvm-project/pull/166558