implement AverageLearner().done()

We now require the user to use goal=lambda l: l.loss() < 1.

Analogous to the IntegratorLearner we can implement learner.done() which tells you whether the tolerance has been reached.