Gather monitoring data for longer time periods

Currently, we only store InfluxDB data for about 7 days, because of limitations with how much data InfluxDB can handle before it stumbles over itself upon startup. This is a problem, because:

• it would be very helpful to see performance data for longer periods than just 7 days
• we are adding more metrics, not taking any away, so there is a danger that if we add more we'll only be able to keep fewer days

My understanding is that we're moving more and more to Prometheus, which doesn't have this limitation: what needs to happen to make this a reality? What are the relevant issues to track? @bjk-gitlab @joshlambert @yorickpeterse ?

\cc @pcarranza