prevent new instances from starting when Cluster.stop() had been issued

[shadeofblue]

closes #461 ?

[johny-b]

This seems to work, although I can't say I really like the solution.
Alternative approach would be to add just a _stopped attribute, without any statemachine.

I think that when we do this sort of a "heavy" change (additional class, additional logic implemented on this class), it would be better to at least have it somewhat complete. E.g. Cluster definitely has a state "stopped", but ClusterState has no such option.

IOW, this is not really Cluster.state, rather Cluster.state_in_few_situations_we_needed_it_and_something_useless_otherwise.

[johny-b]

ALSO: the only reason we need some ClusterState is that we don't have any instance handlers immediately, only after the activity starts. When #560 is done this will be redundant, instead of if self.state == ClusterState.running: we'll have if not instance.stopped:.

And I think that if we could stick to the state of instance only, without state of Cluster, this would be neat.

[johny-b]