Investigate improving scalability of the nanopub pipeline

[Ostrzyciel]

• The pipeline currently caches the individual nanopubs, but not the merge result. We should also cache it.
• The merging is done in rdflib, which will get slow as the dataset expands (see Nanopublications dataset – future scope #19). We could replace it with, e.g., Apache Jena RIOT in CLI mode... or Apache Jena Fuseki, operated via HTTP calls.