BalancingLearner.from_product doesn't work with the DataSaver

The signature is

learner = adaptive.BalancingLearner.from_product(
    f, learner_type=adaptive.AverageLearner, learner_kwargs=dict(atol=0.01), combos=combos)

Currently DataSaver is a wrapper around a learner instance, therefore it's not possible to do this.