Implement tag models

See #432 (closed)

Rather big MR, so let me outline the work done below:

• The first few commits are not directly related to tagging, but introduce a new command, load_test_data that can be used to populate the database with test accounts, tracks, albums, artists, uploads, libraries and tags. This was needed to do basic testing on a database with ~50K tracks and ~100K tags-tracks relationships. Review of the load_test_data.py, associated tests and random-data generators in funkwhale_api/factories.py isn't strictly required, because it's primarily meant for development and won't affect production pods. If you skip that, the MR is actually much smaller ;)
• Dropped django-taggit
• Create a new tags app, which mainly consist of a Tag and TaggedItem model. TaggedItem stores the relationship between tags and other objects (tracks, artists, etc.)
• Added a ?tag filter in the API to test that we can actually filter using tags and validate the performance
• Fixed some issues introduced by the change from taggit.Tag model to our own (e.g in radios)
• Removed obsolete logic because of the changes
• Enable postgres's citext extension to ensure our tags stay unique in DB regardless of the case and make search easier

The MR doesn't include a separate Genre model as initially mentioned in #525 (closed). Before doing that, I'd like to start building the Imorter / UI / APIs / Federation on top of tags, and see how it goes :)

Note: the failing CI test is unrelated (it also affects develop), I'm working on that

cc @funkwhale/reviewers-python