feat: improve triples matching performance

[bishoyroufael]

Hi,

The current triples matching algorithm was an overkill since it needs to retrieve all the triples always. I have improved that by utilizing the correct index function based on what's available. The performance gains are almost 1000x for large graphs in cases like SHACL validation.

Summary:

File changed: rudof_rdf/src/rdf_impl/in_memory_graph.rs

The new triples_matching dispatches to the best oxrdf::Graph index based
on which of the three positions carry a concrete value vs a wildcard (Any):

Pattern	Index used	Complexity
(S, P, O)	graph.contains(triple_ref)	O(1)
(S, P, ?)	graph.objects_for_subject_predicate(s, p)	O(k) — SPO
(S, ?, ?)	graph.triples_for_subject(s)	O(k) — SPO
(?, P, ?)	graph.triples_for_predicate(p)	O(k) — POS
(?, P, O)	graph.subjects_for_predicate_object(p, o)	O(k) — POS
(?, ?, O)	graph.triples_for_object(o)	O(k) — OSP
(S, ?, O)	triples_for_subject(s) + filter by O	O(k_s)
(?, ?, ?)	graph.iter()	O(n) — unavoidable

[samuel-bustamante]

Great optimization, thanks!