Not intuitive use of the energy_axis_computation with dynamic energy binning

[gabemery]

Currently, self.energy_axis_computation is not updated when using the dynamic energy binning.
It means that the value stored in the class can be different from the actually used axis, and that it is harder to access the information without dirrectly passing the axis in function calls.

Proposed fix : change BaseAcceptanceMapCreator._compute_dynamic_energy_axis to update the self.energy_axis_computation instead of returning the axis.

It would also simplify the bin size computation for #77.

[gabemery]

One limit with my solution : the self.energy_axis_computation would be modified in subsequent calls of the class. If this is something we want to avoid, I think we should have a setup function which takes care of handling the reset of changing working attributes.

[mdebony]

I would said the main question is do we want to have an irregular binning set by the user with dynamic energy axis on top of that. If not, I agree with you but we should have clear warning to the users when setting both irregular and dynamic energy axis options.

[gabemery]

One reason to have both together I can see is to have high granularity to create the dynamic energy binning but still provide a lighter final IRF.

[mdebony]

So one possibility to keep this behaviour while still solving your issue above would be to add a third internal energy axis, dedicated to storing the effective computation axis, while retaining the previous ones (maybe with some renaming) in order to keep the current functionalities.

[gabemery]

Branch created new_energy_axis_computation, PR will be made once the fit PR are handled.