Coordinate Onionperf monitoring alerts

Per @hiro's suggestion in irc, this is a ticket to review the Monit configuration we have for onionperf instances, to avoid duplicate checks.

At the moment, monit checks every 5 minutes for:

• whether the onionperf measure process exists
• whether the tgen client process exists
• whether the tgen server process exists
• whether the tor and tgen log files are older than 10 minutes
• whether the disk space is more than 80% used
• whether the instances can reach each other on port 8080

@hiro do we have equivalent checks in prometheus? Do you think we could move to Prometheus and not use Monit?