What are the two conceptual differences between the starter code$$s A$*$ implementation $($for use in CBS$)$ and an A$*$ implementation for single$-$agent pathfinding? We are not looking for implementation details, but for two conceptual changes to the algorithm. That is$,$ CBS would not work without them, while A$*$ for single$-$agent problems would be fine without them. One sentence is enough to describe each difference.