Support for missing data

[mberaha]

When the likelihood is MultiNorm, it should be easy to account for missing data, i.e. observations for which one or more of the entries is missing, by just sampling the missing values from their full conditional distribution.

This should be easy enough, but should require storing the dataset and the indicator of which data are missing.