[llvm] [AMDGPU] Use `v_cvt_pk_*` instructions for i16_f32 saturated conversions (PR #202680)

[Igor Wodiany via llvm-commits]

IgWod wrote:

The PR now only targets f32 -> i16 conversions. I also managed to resolve the issue with GISel so now both GISel and SDAG select packed instructions for both divergent and uniform cases.

https://github.com/llvm/llvm-project/pull/202680