Add inference support (e.g., RDFS)

[donpellegrino]

It would be useful to implement inference support that leverages the advantages of HDT indexes.

The rdflib package supports custom evaluation functions as documented at https://rdflib.readthedocs.io/en/stable/intro_to_sparql.html#custom-evaluation-functions and https://rdflib.readthedocs.io/en/stable/_modules/examples/custom_eval.html#customEval. However, a custom implementation of standard RDFS can be slow and miss opportunities for optimizations that HDT could provide.

Note that Issue #14 and the existing optimize_sparql() may be related.