Properly handle timeouts during ssh commands

[jprider63]

This is from an old email, but I'm posting it here so it isn't lost.

Haskell doesn’t interrupt when running ffi code. This would cause issues when the runner would run a command over ssh (which is implemented via the ffi) and a timeout interrupt wouldn’t fire until the ssh command finished. This would occasionally cause issues where a remote command could hang forever (blocking a team) or the docker image wouldn’t be cleaned up properly. Manually restarting the runner and cleaning up old docker images would fix this temporarily. For the spring contest, I made updates to the ssh library to more gracefully handle interrupts. The issues only happened once during spring 2020, which is an improvement, but there’s still a bug to be tracked down. Rigorously testing this and tracking down this bug would be useful to minimize manual intervention during contests.