Remove hack in __gthread_active_p() to always return 0 on x86_64

This was causing __cxa_guard_acquire to fail and abort if
two threads attempted to initialise the same static local
at same time.

It also presumably had other adverse effects in other
parts of the gcc standard libraries.

TEST= local testing + trybots

Review URL:
diff --git a/gcc/gthr-nacl.h b/gcc/gthr-nacl.h
index ca04410..f99bedd 100644
--- a/gcc/gthr-nacl.h
+++ b/gcc/gthr-nacl.h
@@ -116,15 +116,9 @@
    * TODO: change to reference a thread creation/destruction function
    *             when one becomes available.
-  // This is a dirty workaround over our inability to use locks in NaCl x86-64.
-  // TODO(pasko): fix this by enabling correct builtins.
-#ifndef __x86_64__
   static void *const __gthread_active_ptr
     = __extension__ (void *) &__gthrw_(pthread_mutex_lock);
   return __gthread_active_ptr != 0;
-  return 0;
 #ifdef _LIBOBJC