Add MaskedPrior class for masked variable creation (#1945)

* Add MaskedPrior class for masked variable creation

Introduces the experimental MaskedPrior class to create variables and likelihoods over only the active entries of a boolean mask, expanding back to the full shape with zeros at inactive positions. Includes serialization/deserialization support and comprehensive tests for 1D, multidimensional, and integration with saturation functions and MMM fitting.

* Solving issue with mlflow

* Adding test for Juanito and Codecov

* Adding more tests

* William request

* Applying Juan feedback

* Pre commit

* pre-commit

* Solving issue

* Update special_priors.py

---------

Co-authored-by: Juan Orduz <[email protected]>

Author: cetagostini