[llvm] [AMDGPU] Extend llvm.amdgcn.update.dpp intrinsic to support f64 (PR #91190)

[Matt Arsenault via llvm-commits]

arsenm wrote:

> LGTM, though I wish there was a better way to implement it then duplicating the selection pattern for every 64-bit type.

I've had an open issue for ages to add a type matcher that only cares about the size, but nobody ever gets anywhere with it 

https://github.com/llvm/llvm-project/pull/91190