[clang] [Clang][AMDGPU] Enable type-checking on __builtin_amdgcn_raw_ptr_buffer_atomic_{{add|fadd|fmin|fmax}} (PR #164824)

[Juan Manuel Martinez Caamaño via cfe-commits]

jmmartinez wrote:

> On a side note, one of the the reasons for some compile errors with type check enabling is, we have been using wrong type for FP16. We thought `h` should be fp16 but in fact we should use `x`. There is a PR somewhere else trying to fix it.

It could be the case for `__builtin_amdgcn_raw_ptr_buffer_atomic_fadd_v2f16`.

https://github.com/llvm/llvm-project/pull/164824