bug fixes in computing filtered variances and k-step ahead variances of smoothers

bug fixes in fitered variances of smoother:

• kalman_algo=1 (or 3): k-step ahead variances were WRONG, since Pf was initialized using P in PREVIOUS period.
• kalman_algo=2 (or 4): output argument for filtered variances should be P1, not P (P are UPDATED variances in kalman_algo=2). For kalman_algo=2, I also make a small factorization fix (compute P(:,:,t+1) before defining Pf, so to compute 1-step ahead variance only once).