ZOOKEEPER-3652: Synchronize ClientCnxn outgoing queue flush on a stab…

…le internal value

When packets are added to ClientCnxn's outgoing packet queue we ensure there's no conflict with an ongoing flush of that queue because of connection loss.

Synchronization used to be on the state field's value. This value is both not stable (its value changes over time), possibly causing improper synchronization, and global, which can cause contention in applications that run several ZooKeeper clients.

We now synchronize on outgoingQueue which is both local to a ClientCnxn's instance and stable.

zookeeper-server/src/main/java/org/apache/zookeeper/ClientCnxn.java

@@ -1303,7 +1303,7 @@ public void run() {
                 }
             }
 
-            synchronized (state) {
+            synchronized (outgoingQueue) {
                 // When it comes to this point, it guarantees that later queued
                 // packet to outgoingQueue will be notified of death.
                 cleanup();
@@ -1671,7 +1671,7 @@ public Packet queuePacket(
         // 1. synchronize with the final cleanup() in SendThread.run() to avoid race
         // 2. synchronized against each packet. So if a closeSession packet is added,
         // later packet will be notified.
-        synchronized (state) {
+        synchronized (outgoingQueue) {
             if (!state.isAlive() || closing) {
                 conLossPacket(packet);
             } else {