Support sorting by data source

[hdelva]

Related to #1, if we could add provenance information somewhere to the results, it would be nice to give different weights to the different data sources. This gets interesting because thse weights are probably determined at query-time, so long after the actors have been configured.

We could approach this problem as a reranking problem. Ask the existing actor got 4x more results than we need, and rerank the found results with additional user-determined weights. Or alternatively, we could provide additional parameters to the RdfScore interfaces, but this could risk making all existing actors more complex as well.