Allow connection_fatal to trigger a reconnect attempt, which, if there is
still an issue, which eventually call connection_fatal. This causes us to
retry indefinitely. Maybe eventually we'll need to limit this somehow
* Remove wakeup reconnect delay. Infinite reconnects will fix this itself.
* Make a dns error a normal connection_fatal. Oh dear, oh dear. This was
hardcoded to just messagebox and outright exit(0);. Now we keep
going for a tiny bit, so that we can run the normal cleanup code.
This is the tiny cleanups to ssh.c.