An evaluator function / multiple data, single model

[fjebaker]

It would be very useful to be able to hook into the fitting procedure and overwrite the behaviour when there is say multiple data and only a single model.

• If there's a way to pick out only the normalisations, that would be very good. That way the normalisations can be fit seperately from the rest of the model.

• Add utility functions for taking the union of domains up to some tolerance.

[fjebaker]

See also #132