Add option to have a mean flux parameter per redshift bin

[andreufont]

The current MeanFluxModel parameterizes the effective optical depth ( = exp(-tau_eff) ) with a (log) polynomial multiplying a fiducial evolution (from the central simulation).

We should add an option to model the mean flux with a different parameter per redshift bin.

[andreufont]

This has not been yet implemented, but recently I've added flexibility and tested how the contours degrade when the order of the polynomial goes from 0 to 5, and I find that it saturates at 3... so probably this is not a priority anymore