* Allow customizable ZK update conditions

* Add copyright

* Add unit test, put update condition directly into load manager

This shared reference caused the resource usages between the current report (not published to ZooKeeper) and the
most recently written report to be incorrect and hence when the broker utilization was rapidly changing (e.g., after
a broker restart), the load report only got written once per 15 minutes.

* Fix LoadManager shutdown timing issue

* Remove synchronization on close()

* Remove exceptionally completed topic futures

* Ensure more recently created futures are not removed

* Make candidate selection only consider active brokers

* Fix bugs introduced by new load manager

* Add documentation, improve simulation tools.

* Add global summary for monitor

* monitor -> monitor-brokers

* Don't modify performance producer/consumer

* Use property/cluster instead of cluster/property for simulation

* Fix regex

* handle zkCache failure: invalidate cache and zk-getData failure

* introduce separate executor to serve zkcache callback

* add executor to discovery service

* remove testing npe

* add assetion on zkCache executor parameter and update tests

* update executor thread in testcase for intermittent failure

* Avoid using zk-client thread to process zk-data result

* add ThreadPool-name and stop executors on shutdown

* shutdown zkCache executor if zkCache has initialized it

BrokerService should not remove failed topic from map with the same thread which adds in map to prevent deadlock (#375)