Have robot manager generate_path requests to a queue that multiple worker nodes can process it.
The manager then pulls processed requests and updates the robots as before.

Similarly, as the number of robots increase (100 now, tested at 530 on dev machine successfully) we want to reduce how much info the web visualizer shows

Also, as the warehouse size increases so does the true heuristic dict (24x26 current warehouse ~390 Kb) -> (200x200 would be 1.6 Gb) we can't have each worker node store the true heuristic locally. Consider storing on disk or pulling from redis since it is static.