On costs and benefits of n−2 repetitions in task switching: towards a behavioural marker of cognitive inhibition