A tiny mistake in htlr_predict()

[syumet]

At predict.R line 41 - 48:

if (is.null(usedmc))
{
    if (is.null(burn))
        burn <- floor(no_mcspl * 0.1)
    if (is.null(thin))
        thin <- 1
    usedmc <- seq(burn + 1, no_mcspl, by = thin)
}

If burn == 0 for some reason, then the initial value will be used for prediction (recall that the length of the MCMC sample array is n.iter + 1, where the first entry stores the initial state). So the correct default usedmc should be seq(burn + 2, no_mcspl, by = thin), although it has very little impact on the results.

[syumet]

I will fix it in the next PR.