Provenance runtime/memory optimisations

[alexdunnjpl]

Per #108 (comment)

It should be feasible to block provenance nonaggregate in such a way that a balance between redundant work and unworkable query volumes is achieved. Candidates:

• given that no renamed-product support currently exists, utilize skips such that non-up-to-date products' LIDs are enumerated, then nonaggregates are skipped if they have a LID not in that enumeration.
• one of the core ideals is to query only products having LIDs which are in the set of updated LIDs. An OpenSearch query pattern of "if lid in el1, el2, ... eln" is equivalent in performance to making n individual queries, so is infeasible, given the large quantity of nonaggregate LIDs in existence. But by enumerating the collections containing the affected non-aggregate products and then blocking non-aggregate queries by membership identity, it may be possible to perform efficiently. WIP thought.