Implements the procedure introduced in the paper
Krafty, R. T. and Collinge, W. O. (2013) “Penalized Multivariate Whittle Likelihood for Power Spectrum Estimation.” Biometrika. 100: 447-458.
Computes a smoothing spline based estimator of the power spectrum from a multivariate time series, utilizing the large sample distribution of the periodogram and the Cholesky decomposition to allow for varying levels of smoothness among different components.