Stav dette: Counterfactual choice and learning in a neural network centred on human lateral frontopolar cortex