Enable systemd watchdog for worker

In #937 (comment 642177), Helmut suggested enabling a systemd watchdog for the worker. If the worker were to send keep-alive pings while polling for tasks as well as immediately before and after task execution, we could use the watchdog timeout to enforce a maximum duration for a task.

I haven't fully thought this through, but the idea seems worth splitting out to a separate issue.