Make converting a large table of datasets to RDF faster

[jesper-friis]

In the current implementation each resource is converted separately from JSON-LD to RDF using rdflib. This is slow, both because the JSON-LD implementation in rdflib is slow and because it is very inefficient to call the JSON-LD API for each resource if you are documenting thousands of resources.

Improvements:

• First step: Add a function for combines multiple single-resource dicts into one JSON-LD document with a "@graph"; [...] array in the root containing all the single-resource documents.
• Second step: If it is still slow to document a large table of datasets after the first step, consider to add a method to the backends for uploading a json-ld file to the triplestore (for those triplestores that support json-ld). This may speed up things further, since it is expected that the json-ld implementation in production-ready triplestores like GraphDB is much faster than the rdflib-implementation.