The maximum lag of the fitted model is the maximum lag between ylag and xlag

[wilsonrljr]

The maximum lag of the fitted model should be updated based on the final_model code. If the user saves the model and just want to predict, the initial conditions must have max(xlag, ylag) even if the final model have a lower final lag.

Ex.:

model = PolynomialNarmax(
        ylag=10,
        xlag=10
)

fitted_model = [
    [1001,    0], # y(k-1)
    [2001, 1001], # x1(k-1)y(k-1)
    [2002,    0], # x1(k-2)
    ]

The maximum lag in the version v0.1.6 will be 10. However, setting the maximum lag equal 2 is more intuitive because the user do not have to worry about it.