[AMDGPU] Select no-return atomic intrinsics in tblgen

[abinavpp]

Summary

This is to avoid relying on the post-isel hook.

This change also enable the saddr pattern selection for atomic
intrinsics in GlobalISel.

[arsenm]

llvm/lib/Target/AMDGPU/AMDGPUInstructions.td

558–566

I don't like using int_ prefixes here. This is deceptive and makes it look like you are using raw intrinsics

[abinavpp]

Addressed review comment

[arsenm]

llvm/lib/Target/AMDGPU/AMDGPUInstructions.td

559

This just moved the int_ prefix, and doesn't eliminate it. I just realized these are the cases that only come from the intrinsics, so maybe this is fine (but just directly using int_ before would be better)

[abinavpp]

Revert to Diff 1

[abinavpp]

Rebased

[abinavpp]

Ping; This is required for D124086