Modulation of temporal decision-making in a deep reinforcement learning agent under the dual-task paradigm