Expose the mean model for CBV correction to the user

[mirca]

Sir @barentsen mentioned today that one could jointly fit for the CBV weights and an astrophysical variability (so that the astrophysical signal wouldn't be taken out by the CBVs) by exposing the mean_model which is hard coded here

[mirca]

For instance, for EPIC248667471, we can see that the CBVs degraded the lightcurve most likely because of the lack of knowledge about the periodical signal.