Do not re-backup archived repos

Following out of #3 which is mostly implemented. The terminology here is about to suck. Strap in.

This probably needs to rely on some kind of local database as provided by #10. If a project is newly marked as archived then we need to download its archive. If you unmark a project as archived we need to start downloading its archives again. If the old download was for an archive project and the project is still marked as archived then we don't need to redownload its archive. While this _could_, in theory, miss changes if you unarchive, change, and re-archive, I don't think that's a super-useful case to handle (unless we can get timestamps from GitHub).

Start by reviewing the implementation referenced by #3 and the local-database proposal in #10. Define the archived-state transitions described here: download newly archived repositories once, avoid repeating downloads while they remain archived, and resume downloads after unarchiving.

Written by the indexing model from the issue text.