ZOOKEEPER-1990: fix Random instances

See the jira for more info.

Basically we have multiple ways of creating Random instances in ZooKeeper. Since java 1.7, the default constructor is good enough even in multi-threaded environment, we get a good seed.
But in some places, we just create a random instance, where System.nanotime is the seed, which is not a good practice in multi-threaded environments.

I only replaced those, and I also left the tests as is, because in some cases it is intentional in them.

I created the PR to bring more attention to the ticket, please feel free to share your ideas on the topic!

Author: Norbert Kalmar <nkalmar@yahoo.com>

Reviewers: fangmin@apache.org, andor@apache.org

Closes #617 from nkalmar/ZOOKEEPER-1990

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

@@ -801,7 +801,7 @@ public class ClientCnxn {
     class SendThread extends ZooKeeperThread {
         private long lastPingSentNs;
         private final ClientCnxnSocket clientCnxnSocket;
-        private Random r = new Random(System.nanoTime());        
+        private Random r = new Random();
         private boolean isFirstConnect = true;
 
         void readResponse(ByteBuffer incomingBuffer) throws IOException {

src/java/main/org/apache/zookeeper/server/SyncRequestProcessor.java

@@ -61,7 +61,7 @@ public class SyncRequestProcessor extends ZooKeeperCriticalThread implements
      * invoked after flush returns successfully.
      */
     private final LinkedList<Request> toFlush = new LinkedList<Request>();
-    private final Random r = new Random(System.nanoTime());
+    private final Random r = new Random();
     /**
      * The number of log entries to log before starting a snapshot
      */