[clang] [llvm] [NVPTX] Add errors for incorrect CUDA addrpaces (PR #138706)

[Lewis Crawford via llvm-commits]

LewisCrawford wrote:

I think it might be best to just close this, and leave it to front-ends to validate instead of adding IR-level validation, as it would make certain IR-level tests a lot more inconvenient to write. The chances of someone hitting problems from using invalid address-spaces like this in a real kernel in a way that bypasses any error-checking the front-end are a lot lower than someone hitting the new error message from this patch while trying to write an IR-level test and having to find an inconvenient workaround for it.

https://github.com/llvm/llvm-project/pull/138706