manager: Return early from pipeline instantation if any device failed to instantiate
Nothing seems to worry the manager for long, seeing it happily switches to ACTIVE from ERROR after instantation and clean-up is done. I'm afraid we need to dwell on such setbacks for a bit.