Use separate sets to cache entity and relationship keys

[austinkelleher]

I don't think this is a very common use case, but an entity key and a relationship key can technically be the same and we are currently using the same Set (DuplicateKeyTracker) to track the keys for each state.

[aiwilliams]

@philidem hadn't we thought to pursue key uniqueness even across entities and relationships? Or, are we forever stuck with a compound key? What is the compound key then? It must not be ['_type', '_key'] or else we could allow duplicate keys across entities as long as their _type is different.