In support of T310598.

Let's audit all existing alerts and on-call schedules with the following goals in mind:

1. Improve paging by prioritizing team members who are awake
2. Prioritize alerts to ensure pages only occur for critical events.
3. Remove unneeded alerts. Move informational alerts to automated tickets, rather than pages via https://phabricator.wikimedia.org/p/phaultfinder/
4. Make alertmanager the single source of truth and interface for visualizing and responding to alerts