[llvm] [AMDGPU] Add OMOD folding for TRANS bfloat16 instructions (PR #208595)

[Jay Foad via llvm-commits]

jayfoad wrote:

> I think they should belong to the same PR. however, let me know if you still think otherwise, so I can split it. Thanks;

You haven't said anything to change my mind.

The reason for splitting is to simplify the review, so you can hopefully get the simpler V_FMA_MIX* part committed first, and then we can iterate on the V_CVT_PK_BF16_F32 part. The latter is more complicated because you need to define what `isOMod` means on an instruction with a vector result, and why your definition makes sense.

https://github.com/llvm/llvm-project/pull/208595