* Fix a few issues

* Make sure that adding a dispatch return event to a node stream creates the stream if needed.
  This could cause jobs that are added right on startup to fail to be dispatched.
* Fix potential panic in cleanup goroutine when the pending jobs map contains nil values
* Make sure to close all maps when closing the node.
* Properly log event information when discarding stale ack events.
* More consistent logging

* Fix a few additional issues

* Ensure stale node streams are gced
* Make close and shutdown code cleaner
* Cleanup all maps properly on node close