Generalize the reward function to allow heuristics from the simulator

[ancorso]

Allow the reward function to accept heuristics from the simulator natively (these are quantities that guide the search algorithm like how the close the vehicle was to a collision which)