[llvm] [AMDGPU] Add intrinsic for raw atomic buffer loads (PR #97707)

[Jessica Del via llvm-commits]

OutOfCache wrote:

> Could you go ahead and address the TODO at https://github.com/llvm/llvm-project/blob/main/llvm/lib/Target/AMDGPU/AMDGPULowerBufferFatPointers.cpp#L1095-L1096 ? That's left over from this intrinsic not being upstreamed

I believe I did in the last fixup commit `93b6b5b776667997fcea77ec9b52c72c2b5b3ff5`? The lit test also shows the use of the atomic buffer load.

I removed the addition in InstCombine so we can add them in a separate PR with an `opt` test, as requested. 

https://github.com/llvm/llvm-project/pull/97707