Gravar-mail: Time representation in reinforcement learning models of the basal ganglia