Improved mcse_elpd_loo

Currently, mcse_elpd_loo is computed by assuming the pointwise mcse's to be independent. As the same MC draws are used for each pointwise PSIS weights, the weights are correlated and this correlation can sometimes to be high enough that it would be useful to take into account. The extension of the formulas is straight forward, but as the current approach can compute pointwise mcse's parallely, extra care is needed to check that all different functions would work given the change, or at least information would be provided whether independence assumption has been used or not. I'll prepare a small case study to illustrate computation for improved mcse.