Currently when invoking the sampling estimator on index scan nodes, we pass the equality prefix of the bounds as a heuristic to account for the additional seeks that the cursor must perform. For example the bounds of [a: [1,1], b: [2, 5], c: [6,6]] are estimated as [a: [1,1], b: [2, 5], c: [MinKey,MaxKey]]. Our current approach creates a copy of the index bounds to do this as a convinence, but this is unnecessary.

The task of this ticket is to avoid these allocations.