SparqlQuerySelector should take into account duplicate quads

[pmaria]

For SPARQL endpoints that do not deduplicate constructed quads, given a dcat:Dataset description with multiple distributions, the SparqlQuerySelector will currently construct a dataset object per distribution instead of correctly constructing a single dataset object with multiple distributions, matching the source data.

To account for this the SparqlQuerySelector should take into account duplicate quads and should keep track of processed datasets.

See also:

• comunica-sparql-file: CONSTRUCT query returns duplicate triples comunica/comunica#1105
• Duplicate bindings on construct query comunica/comunica#1300

[ddeboer]

Looks like deduplication will be solved in Comunica: comunica/comunica#1388.