Have a process to ensure squashing DB migrations are done regularly

Context

In Investigate db:check-schema slowdown (#564 - closed), we noticed that the DB migrations weren't squashed for a year.

Improve migration squash tooling (gitlab-org/gitlab#498538 - closed) should greatly simplify the DB migrations squashing process. We would still need to ensure this process is done regularly (ideally by groupdatabase).

Goal(s)

• Have a process we could follow in Engineering Productivity to ping the DRIs to squash DB migrations (or, if the process really is straightforward, just do it ourselves on a regular basis).
• Ensure the process to do generate a squashing MR is documented (even if it ends up being a one-liner like "MRs are generated automatically every X months, and are assigned to groupdatabase for review"

Things to figure out

• Smallest interval at which we could safely squash DB migrations (every release, every 3 releases, ...). The smaller the number of migrations, the better.