Preemptive Solving of Future Problems: Multitask Preplay in Humans and Machines