Gossip merge in large cluster, #2290

* Trying to simultaneously resolving conflicts at several nodes creates new conflicts.
  Therefore the leader resolves conflicts to limit divergence. To avoid overload there
  is also a configurable rate limit of how many conflicts that are handled by second.
* Netty blocks when sending to broken connections. ClusterHeartbeatSender actor
  isolates sending to different nodes by using child workers for each target
  address and thereby reduce the risk of irregular heartbeats to healty
  nodes due to broken connections to other nodes.

akka-cluster/src/test/scala/akka/cluster/ClusterConfigSpec.scala

@@ -35,8 +35,13 @@ class ClusterConfigSpec extends AkkaSpec {
       AutoDown must be(true)
       UseDispatcher must be(Dispatchers.DefaultDispatcherId)
       GossipDifferentViewProbability must be(0.8 plusOrMinus 0.0001)
+      MaxGossipMergeRate must be(5.0 plusOrMinus 0.0001)
       SchedulerTickDuration must be(33 millis)
       SchedulerTicksPerWheel must be(512)
+      SendCircuitBreakerSettings must be(CircuitBreakerSettings(
+        maxFailures = 3,
+        callTimeout = 2 seconds,
+        resetTimeout = 30 seconds))
     }
   }
 }