[llvm] [BPF] Report an error if comparison imm operand cannot fit in 32bit (PR #142989)

[via llvm-commits]

yonghong-song wrote:

> LGTM, but what is the conclusion of the upstream discussion? should the error be reproted for `Imm > UINT_MAX` or for `Imm > INT_MAX`? I think the latter makes sense, as you suggested upsteream.

Yes, I do prefer Imm > INT_MAX. But I will emit warning instead of error to maintain some kind of backward compatibility.

https://github.com/llvm/llvm-project/pull/142989