Implements the Butterworth-Induced Autoregressive ('BTWAR')
model, where autoregressive coefficients are obtained from analog
Butterworth filter prototypes mapped into the discrete-time domain
using the Matched Z-Transform. The framework establishes a structured
connection between frequency-domain filter design and time-domain
autoregressive modeling. Model order selection is performed via
nested rolling-origin cross-validation. Method described in
Bras-Geraldes, Rocha and Martins (2026)