Dedup ScheduleStream by jobName.

[debugger87]

chronos with only one server instance will be switched to Defeated status if it's disconnected to zookeeper and be switched to Elected if it's connected to zookeeper. In this case, chronos load job from zookeeper and add ScheduleStream into streams in JobScheduler. If we don't dedup ScheduleStream by jobName, the size of streams will increase unexpetedly as follows:

[INFO] [pool-5-thread-1] 06-10 00:01:00,468 [JobScheduler] - Size of streams: 150
......
[INFO] [pool-5-thread-1] 06-10 19:01:46,113 [JobScheduler] - Defeated. Not the current leader.
[INFO] [main-EventThread] 06-10 19:01:59,900 [ConnectionStateManager] - State change: RECONNECTED
[INFO] [pool-5-thread-1] 06-10 19:01:59,905 [JobScheduler] - Elected as leader.
....
[INFO] [pool-5-thread-1] 06-10 19:02:35,965 [JobScheduler] - Size of streams: 231

Obviously, there are many ScheduleStreams which are duplicated in streams. If some jobs' tasks fail, chronos' JobScheduler may run into unexpected status like dead lock.

[debugger87]

@brndnmtthws could you help me to review this PR？

[brndnmtthws]

Interesting. Thanks for the PR. Can you describe how you tested it?

[debugger87]

@brndnmtthws We can just launch only one server instance of chronos, shut down and restart zookeeper in some time (make chronos disconnect to zookeeper and re-elected as leader). In this case, streams in JobScheduler will append duplicated ScheduleStreams.