Upon #storage reaching #autostart, with autostart being on, by definition if recovery / verification phases can be completed, the cluster should change its state to RUNNING.

The corresponding test (introduced in 58774fb6 "master: new option to automatically start a new cluster") was not checking that.