[llvm] [AMDGPU][wmma] - Add tied wmma intrinsic (PR #69903)

[Nicolai Hähnle via llvm-commits]

nhaehnle wrote:

Thanks. Can you please still add a comment in IntrinsicsAMDGPU.td that for f16 / bf16 versions, the tied version copies the "other" half of the accumulator to the result, while in the non-tied versions, the "other" half of the result is undefined?

https://github.com/llvm/llvm-project/pull/69903