Make ResourceWatcher responsible for raising group-status-based errors

[KnVerey]

Currently the RunnerTask, RestartTask and DeployTask each have some code like this:

      raise DeploymentTimeoutError if resources.any(&:deploy_timed_out?)
      raise FatalDeploymentError if resources.any?(&:deploy_failed?)

We should see if it could work to move that responsibility to the ResourceWatcher to avoid the duplication. Note that the watcher is already responsible for raising global timeout errors.

[dturn]

I have mixed feelings on if the resource watcher should raise or not. But I definitely think resource watcher should be responsible for checking the overall status of a deployment instead of checking each resource individually. I started down that route in the state machine refactor #303