Inference performance improvements about dannet

[simongray]

Once other complication is that copying the inferred triples from an InfModel to a TDB graph takes up a great deal of memory, e.g.

(def persisted-inf
  (-> dannet
      :db-path "resources/tdb2-inf"
      :db-type :tdb2))

(q/transact (:model persisted-inf)
  (.add (:model persisted-inf) model))

Attempting the above using the default (I think 4GB) heap space results in a an out of memory error when the model uses OWL micro inferencing.

This will need to be resolved, so that it becomes possible to infer all possible triples of some snapshot of the base graph and populate a second graph - likely a TBD, to ensure persistence - with all of the inferred data.

[simongray]

Increasing to 8GB max heap space also yielded an OutOfMemoryError when copying all triples from an InfModel to an empty TDB 2 model. It seems that unless the sweet spot is slightly above 8GB, preserving all of the triples inferred by OntModelSpec/OWL_MEM_MICRO_RULE_INF is simply out of the question.

I also tried the operation using an RDFS inference model and this completes in a very short time, likely due to little-to-no actual inference taking place. Unfortunately, there is no ready-made OntModelSpec that only infers opposite relations which is the primary need I have.

[simongray]

I initially implemented a custom Reasoner based on the OWLMicroReasoner class with a custom .rules file and closed the issue in that commit, but it turns out that no Java code is necessary, as one can simply instantiate a GenericRuleReasoner to apply the same custom rule set: fb7329d