[clang] [AMDGPU] add macro `__AMDGCN_CDNA_VERSION__` (PR #88293)

[Joseph Huber via cfe-commits]

https://github.com/jhuber6 commented:

Seems somewhat superfluous to me. I'm wondering if what this customer wants to do is more easily serviced by `#if __has_builtin(...)` or `#if __has_feature(...)`. However I definitely agree that our feature detection macros are a bit of a mess. I'm not privy to the naming methodology here, presumably future generations will not be called `CDNAx` so wouldn't be then have some special case code for one family that doesn't expose any extra information?

https://github.com/llvm/llvm-project/pull/88293