streamline mutex unlock to remove a useless branch, use a_store to unlock

this roughly halves the cost of pthread_mutex_unlock, at least for
non-robust, normal-type mutexes.

the a_store change is in preparation for future support of archs which
require a memory barrier or special atomic store operation, and also
should prevent the possibility of the compiler misordering writes.

src/thread/pthread_mutex_unlock.c

@@ -14,11 +14,15 @@ int pthread_mutex_unlock(pthread_mutex_t *m)
 			self->robust_list.pending = &m->_m_next;
 			*(void **)m->_m_prev = m->_m_next;
 			if (m->_m_next) ((void **)m->_m_next)[-1] = m->_m_prev;
+			a_store(&m->_m_lock, 0);
+			self->robust_list.pending = 0;
+		} else {
+			a_store(&m->_m_lock, 0);
 		}
+	} else {
+		a_store(&m->_m_lock, 0);
 	}
 
-	m->_m_lock = 0;
 	if (m->_m_waiters) __wake(&m->_m_lock, 1, 0);
-	if (m->_m_type >= 4) self->robust_list.pending = 0;
 	return 0;
 }