Adjust to the new clang builtin modules

<__stddef_size_t.h> is now a modular header, and so will not be absorbed into Darwin. When that happens, Darwin.C.stddef no longer has a size_t declaration, instead it imports the declaration from _Builtin_stddef.size_t. Darwin still has its other size_t declaration though, Darwin.POSIX._types._size_t, and that's the one that will be found currently. If that one ever goes away though, allow for _Builtin_stddef.size_t being found in C-typedef-Darwin.

Author: ian-twilightcoder

test/DebugInfo/C-typedef-Darwin.swift

@@ -9,4 +9,12 @@ let blah = size_t(1024)
 use(blah)
 // CHECK: !DIDerivedType(tag: DW_TAG_typedef,
 // CHECK-SAME:           scope: ![[DARWIN_MODULE:[0-9]+]],
-// CHECK: ![[DARWIN_MODULE]] = !DIModule({{.*}}, name: "stddef"
+// size_t is defined in clang by <__stddef_size_t.h> (_Builtin_stddef.size_t)
+// and in macOS by <sys/_types/_size_t.h> (Darwin.POSIX._types._size_t). When
+// Darwin builds, Darwin.C.stddef will precompile clang's <stddef.h>, which will
+// include <__stddef_size_t.h> so that declaration will be visible via the
+// `export *`. Darwin.POSIX._types._size_t will precompile <sys/_types/_size_t.h>
+// into the Darwin pcm, making it the likely source of size_t. However, if Apple
+// stops redeclaring size_t, then the clang one will be the only one, so check
+// for either module.
+// CHECK: ![[DARWIN_MODULE]] = !DIModule({{.*}}, name: "{{_?size_t}}"