SDL_memcpy.c, SDL_memmove.c, SDL_memset.c: don't use gcc builtins if !HAVE_LIBC

__builtin_memcpy, as well as __builtin_memset and __builtin_memmove, needn't be
inlined but emitted as a libc call, leading to infinitely recursive calls.

Fixes https://github.com/libsdl-org/SDL/issues/9090

src/stdlib/SDL_memcpy.c

@@ -29,7 +29,7 @@
 #endif
 void *SDL_memcpy(SDL_OUT_BYTECAP(len) void *dst, SDL_IN_BYTECAP(len) const void *src, size_t len)
 {
-#ifdef __GNUC__
+#if defined(__GNUC__) && (defined(HAVE_LIBC) && HAVE_LIBC)
     /* Presumably this is well tuned for speed.
        On my machine this is twice as fast as the C code below.
      */
@@ -76,7 +76,7 @@ void *SDL_memcpy(SDL_OUT_BYTECAP(len) void *dst, SDL_IN_BYTECAP(len) const void
         }
     }
     return dst;
-#endif /* __GNUC__ */
+#endif /* HAVE_MEMCPY */
 }
 
 /* The optimizer on Visual Studio 2005 and later generates memcpy() and memset() calls.

src/stdlib/SDL_memmove.c

@@ -29,7 +29,7 @@
 #endif
 void *SDL_memmove(SDL_OUT_BYTECAP(len) void *dst, SDL_IN_BYTECAP(len) const void *src, size_t len)
 {
-#ifdef __GNUC__
+#if defined(__GNUC__) && (defined(HAVE_LIBC) && HAVE_LIBC)
     /* Presumably this is well tuned for speed. */
     return __builtin_memmove(dst, src, len);
 #elif defined(HAVE_MEMMOVE)

src/stdlib/SDL_memset.c

@@ -29,7 +29,7 @@
 #endif
 void *SDL_memset(SDL_OUT_BYTECAP(len) void *dst, int c, size_t len)
 {
-#ifdef __GNUC__
+#if defined(__GNUC__) && (defined(HAVE_LIBC) && HAVE_LIBC)
     return __builtin_memset(dst, c, len);
 #elif defined(HAVE_MEMSET)
     return memset(dst, c, len);