Time-varying system matrices for non-uniform timesteps.
Observation vector [m] (row of F for p=1) — same for all steps
Per-step transition matrices G(Δt_k) — shape [n, m, m] as nested JS arrays
State dimension
Per-step noise covariances W(Δt_k) — shape [n, m, m] as nested JS arrays
Time-varying system matrices for non-uniform timesteps.