Optimize the repos API query

[davidfischer]

The goal of this PR is to speed up the https://readthedocs.org/api/v2/remote/repo/ API endpoint. Specifically this is used on the repository import page and we've seen some timeouts.

• Adds an index to Project.repo. This is queried heavily and without an index we are looking at a full table scan. This is useful not just here but elsewhere as well.
• There is a straight forward select_related optimization on repository organizations
• In order to disallow importing duplicate repos, we currently do a rather inefficient fuzzy match on repositories. This query is done for each repository in the query resulting in up to 15 additional queries (each of which currently results in a full table scan). This change attempts to optimize this matching query but it isn't totally lossless:
  • The change will no longer find case insensitive matches
  • The query attempts to do a very similar fuzzy match with respect to git:// URLs or URLs ending in .git but it is not exactly the same

Fixes #5441

[ericholscher]

Looks good to me. 👍

[webknjaz]

Hi, when do you expect to deploy this?

[ericholscher]

Should be deployed early this week.