Text this: A task-invariant prior explains trial-by-trial active avoidance behaviour across gain and loss tasks