Clean up dead event directories automatically

[jcnelson]

We're going to run out of space on /dev/events and /dev/metadata/udev/events if we run for too long, since a process can die without cleaning itself up. We'll need to either use runfs directly, create eventfs from runfs (optimized for deque-like access patterns; see [1]), or create a "reaper" process that runs every X minutes to clean up dead event directories.

[1] http://comments.gmane.org/gmane.linux.devuan.devel/2888

[jcnelson]

See https://github.com/jcnelson/eventfs. Currently a WIP, but will be suitable once ready.

[fbt]

Wow. eventfs looks really cool. Got nothing constructive to contribute, just wanted to say that.