[clang] [llvm] [AMDGPU] Add v_cvt_scale_pk8_* gfx1250 instructions (PR #151616)

[Stanislav Mekhanoshin via llvm-commits]

rampitec wrote:

Literally the only viable way to commit is to 'bypass rules' now.

https://github.com/llvm/llvm-project/pull/151616