[clang] [clang][AMDGPU] Clean-up handling of named barrier type (PR #207687)

[Jay Foad via cfe-commits]

jayfoad wrote:

> > Instead of just disallowing inheritcance can you impose a standard restriction like std::is_standard_layout?
> 
> If there is a need for it, I can do it. I looked into it briefly, it adds a bit of complexity to the diagnostics. Nothing too bad but I need to rewrite most of the Sema logic again. I think to do this right, I need to propagate the new Attribute throughout the entire class hierarchy. I will need a bit of time to get that up and running with proper test coverage and diagnostics.
> 
> Let me know if it's really needed or not and I'll get to work

I can't say that it's required, it was just an idea to use the (hopefully) correct C++ concept for this instead of an ad hoc check for inheritance.

https://github.com/llvm/llvm-project/pull/207687