database/repos: More memory efficent private repos observation

Created by: tsenart

I noticed we were aggregating a list of IDs unnecessarily. This is a small optimization in a hot code path.