purge old containers / images / volumes / networks when docker-engine startup

[jonasfj]

When the worker startup we should purge all old containers, images, volumes and networks that might have been left-over from previous executions, in case the worker previously crashed.

I propose we either (or both):
A) tag all resources with "owner": "taskcluster-worker" and any such tagged resources at startup.
B) make this a worker configuration option, that is disabled by default.

This would avoid us running out of disk space after a crash where resources were leaked.