Added support for the KFAS backend for state-space filtering and smoothing routines. The backend can be selected using the package argument (package = "KFAS").
This backend is particularly useful when the forecast error variance matrix is singular or near-singular, a situation in which the default FKF implementation may fail. While KFAS provides a more robust alternative, it is computationally more demanding and may result in longer execution times.