Merge EDB data from several sources if possible at low cost

[mkroetzsch]

Currently, every EDB predicate can only be based on one data source. However, all file-based inputs (n3, csv) as well as in-memory facts declared directly are presumably managed in the same in-memory data structure. It would be useful if VLog could simply add facts from several such sources to the same EDB predicate rather than requiring each input to use a different predicate. For example, it would be good if several n3 files or several csv files could be loaded into the same predicate.

DBMS-based inputs or SPARQL sources should probably not be merged in this way (since this would incur significant cost), but that's fine.