Re-purposing Compact Neuronal Circuit Policies to Govern Reinforcement Learning Tasks