Feature: Restart the piston process if it dies unexpectedly

[rehanift]

Right now if the piston process crashes (ie. segfaults) there is no logic to restart the process and continue processing tasks. This leaves the system in a bad state (ie. incomplete tasks orphaned on cylinders).

Additionally it would be nice to log when these events (crash & restart) happen and during which task execution

[ericallam]

Any idea on what kind of work would be involved in getting this through? I'm seeing periodic piston failures (usually timeout errors) and then no more tasks are able to be executed until the server is restarted.