monitoring: Alert on massively bloated tables

Created by: efritz

We have monitoring for postgres instances, but what's on the dashboard is pretty light. We've had recent recent customer issues that caused outages due to high table bloat (and some failure of autovacuum, which we will root-cause separately).

We should be monitoring some health (or proxy of health) of the autovacuum/autoanalyze daemons in these instances.