Allow more leeway in time between uptime checks running to avoid falsely reporting a service has gone down

The uptime monitoring service (Stackdriver) seems to occasionally leave 60s gaps between running an uptime check. This currently causes our alerts to fire because we expect at least one uptime check to run every 60s.

We should increase the time in which we expect to see an uptime check run, and additionally look at explicitly looking for uptime check failures in order to allow us to be notified quickly if a check has failed.