Improved annotation/assignment handling

we probably need a (project-level) way to define "relevance" (e.g. via stored search)
Quality monitoring
- level of agreement per label per scope (and/or scheme)
- Cumulative stats (e.g. number of relevant docs)
Label resolution
- Take out resolution algorithms that are not implemented
- Resolve labels for documents that no-one annotated (Label centre)
- New resolution algorithm: only resolve where people agree, otherwise leave empty
Assignments
- more options beyond random sample (+ excl. prior scopes)
- ideally via (stored) internal search queries