Selective sync at gitlab.com scale

For the migration project we are looking into ways to make it possible to selectively sync some repositories or groups so we can start validating geo with production traffic before turning it on for all repositories, possibly creating a large amount of load on the primary.

As it stands currently the selective sync feature does not work at gitlab.com scale as it doesn't perform well because it is not using FDW for queries.

Selecting by storage shard would be the most beneficial to gitlab.com as it would allow us to incrementally bring storage online, slowly increasing the number of repositories that are scheduled to be sync'd to the secondary.