In the future, we need to refactor our perturbation matrix API.

• A, A_cp, A2, and A3 are impossible to understand when reading the code
• We ideally need some sort of perturbation matrix class which deals with the building and shaping of this high dimensional matrix
• This would let us append other vectors (e.g. engineering data)
• This class might have methods for "cleaning" the input vectors (detrending by polynomials, fitting with splines to make smooth etc)
• This would let us have some function like mac.mean_perturbed_model(time=...) which would give us the flux from the perturbed model!

This is currently NOT a priority.