Add missing storeload memory fences

There is no synchronize with relationship between the monitor micro-
lock and the onDeck microlock, so it is possible for an onDeck.load to
move above a contendersList.store, or a contendersList.load to move
above an ondeck.store.

To fix this issue a full memory fence (mm_mfence on x86) is needed
after the last store in the contendersList and onDeck critical regions.

Change-Id: I5beb7dfe0d21010c5bf00cd65d59b9c7af58e919

rocclr/thread/monitor.cpp

@@ -205,6 +205,10 @@ void Monitor::finishUnlock() {
       return;
     }
 
+    // A StoreLoad barrier is required to make sure the onDeck_ store is published before
+    // the contendersList_ micro-lock check.
+    std::atomic_thread_fence(std::memory_order_seq_cst);
+
     // We do not have an on-deck thread (semaphore == NULL). Return if
     // the contention list is empty or if the lock got acquired again.
     head = contendersList_;