Make ownership filter consistent between relink and rimport

As of #31, `rimport` walks directories to find files to publish with no regard for who owns the file. `relink.py`, on the other hand, still filters to just the owner. This is necessary, I think, because the `rimport` makes it so commands are run as the staging owner, and thus we would need special handling to check that ownership by _the calling user_ is checked.

It may be acceptable to go either way on this: Both check ownership or neither does. It feels weird to have this obvious failure case possibly lead to situations where someone publishes someone else's file but then can't relink it, so I lean towards both checking.

Compare the ownership filtering in relink.py with the directory walk in rimport, focusing on how the staging owner and calling user are represented. Resolve whether both commands should check ownership or neither should, then verify that publishing and relinking handle another user's file consistently.

Vom Indexierungsmodell aus dem Issue-Text verfasst.