[llvm] [AMDGPU] Folding imm offset in more cases for scratch access (PR #70634)

[via llvm-commits]

ruiling wrote:

> > I don't know. The whole `CheckBothOperands` thing seems fragile to me and as if something wasn't quite right yet. See also the inline comments.
> 
> `CheckBothOperands` is for the case when we put both operands into SGPR and VGPR separately, so we need to check against both operands. If the flag is false, we only need to check against the first operand which will be put in SGPR/VGPR. Feel free to propose a better way for this:)

The `CheckBothOperands` has been removed in latest commit.

https://github.com/llvm/llvm-project/pull/70634