[clang] [AMDGPU] Add clang builtin for generic AMDGPU shuffle (PR #185302)

[Shilei Tian via cfe-commits]

================
@@ -213,6 +213,7 @@ def __builtin_amdgcn_ds_permute : AMDGPUBuiltin<"int(int, int)", [Const]>;
 def __builtin_amdgcn_ds_bpermute : AMDGPUBuiltin<"int(int, int)", [Const]>;
 def __builtin_amdgcn_readfirstlane : AMDGPUBuiltin<"int(int)", [Const]>;
 def __builtin_amdgcn_readlane : AMDGPUBuiltin<"int(int, int)", [Const]>;
+def __builtin_amdgcn_wave_shuffle : AMDGPUBuiltin<"int(int, int)", [Const]> { let Documentation = [DocWaveShuffle]; }
----------------
shiltian wrote:

Use `ArgNames` here such that you don't need to add the prototype in the documentation content but can still use argument names.

https://github.com/llvm/llvm-project/pull/185302