RSC should timeout and exit if Livy doesn't connect back

After we reversed the way that Livy and RSC connections are established, the RSC may now listen forever for a Livy connection that may not arrive. This would leave the session using resources in the cluster and would require manual intervention from users to kill the application.

Instead, there should be a timeout: if there are no connections to the RSC after a configurable x amount of time, it just stops itself.