Incorrect hidden args setting for ensembles

[paxcema]

Something like this:

pdef = ProblemDefinition.from_dict({'target': target, 'time_aim': 80})
jai = json_ai_from_problem(df, pdef)
jai.model['module'] = 'MeanEnsemble'

predictor = predictor_from_json_ai(jai)
predictor.learn(df)

Will currently fail because the autogenerated call to the ensemble has arguments incorrectly ordered (and some of them may not be required).