If a search in a spatial partition is taking place entirely within a single bin, the cost of tracking possible duplicate items with a set can be avoided for a small speedup.
Reduce code duplication by extracting a common class to deal with spatial partitioning of actors, and use some (cached) delegates to reduce duplication further without affecting performance too much.
Speed up updates and removal of actors by caching their location so we only need to update or remove them from bins they are actually in (typically very few), compared to having to check every bin for removals which is much more work in comparison.
Speed up checking for actors inside a region by checking if items are located entirely within the bin they are located in. If so, we don't need to add them to the hash-set for de-duplication purposes which is fairly expensive.
