[llvm] [AMDGPU] Disallow null for more resource operands (PR #121941)

[Jay Foad via llvm-commits]

jayfoad wrote:

I found these by inspecting the tablegen output, looking for InOperandLists that used e.g. SReg_128 instead of SReg_128_XNULL.

There are still lots of selection patterns that refer to the non-_XNULL classes. I don't know if there's any need to update those too.

https://github.com/llvm/llvm-project/pull/121941