[clang] [Parser][NFC] Move the core parsing of an attribute into a separate method (PR #107300)

[Erich Keane via cfe-commits]

================
@@ -2943,6 +2943,9 @@ class Parser : public CodeCompletionHandler {
     return false;
   }
 
+  bool ParseGNUSingleAttribute(ParsedAttributes &Attrs, SourceLocation &EndLoc,
----------------
erichkeane wrote:

Ah, I see.  The WG14 specified one is `__builtin_has_c_attribute`, correct?  

That said, I'd think that we would want the Clang behavior to answer the right thing for ALL of our attributes (including the MSVC ones).  But I then wonder if, since this is a GCC originating builtin, if we should only accept ones with a GCC spelling.

So I guess I'm on the fence here.  I REALLY wish the name was more GCC specific, that way we could just match their behavior, but with the name as it is, I think we have to answer that question for ANY attribute, including declspecs, standard attributes, or clang extensions.

So perhaps that changes this refactor?

https://github.com/llvm/llvm-project/pull/107300