The current implementation of `UpdateCoordinatorStandard` leads to an high amount of allocations, as it attempts to track which tables are going to be updated by using 4 different HashSet implementations.

These 4 HashSet instances are being allocated for each update operation, and each entity instance within the overall update; some of them will remain empty.

Also noticeable, each Table within this scope is identified by a unique, short number. I’m planning to introduce a small and simple wrapper around BitSet to replace these sets.