scanner: always determine known content (before looking for origin)

The known request is significantly cheaper and simpler to run. Especially, it can be run in any order, allowing for more efficient paraller fetching.

The origing request as we want to run request as high as possible in the source tree, to have more accurate result.

Even when having to do fuzzier matching in the future, known which neighbour are locally known will help having better origin answers.

So lets perform the known queries early in all cases.