allow quadtree to work w/ overlapping positions

[kenblu24]

If two agents have the exact same position in the world, then inserting them into the quadtree will fail with RecursionError: maximum recursion depth exceeded.

This PR makes several small changes to how we use quadtrees:

• The quadtree now stores a list of agents at each point
• We first query the quadtree prior to insertion, and if the same point is found, append the new agent to the list.
• There is now a fallback for the BinaryFOVSensor in the case that world.quad is somehow not found
• I now use numpy arrays instead of withinRadiusExclusiveFast to filter the consideration set by sensor radius

This code was tested using a milling setup with 6-100 agents, as well as with the docker orb simulator (SITLUTMWorld, not in this repo)

[kenblu24]

@jhudd135 If you're still working on the sim, you may want to pull main into your branch to incorporate #38 and #40 which affect the BinaryFOVSensor.