Allow for nuisance parameters

[LucaMantani]

We should allow for the possibility to include nuisance parameters, e.g. we could fit the Lambda positivity and integrability.

[LucaMantani]

We could probably start with the simplest option, a multiplicative and an additive nuisance:
pred = T * f(c) + r(c')

Not sure how simply it could be to have the functions f and r rather generic, but that would be really flexible.