Improved robustness of reinforcement learning policies upon conversion to spiking neuronal network platforms applied to Atari Breakout game.