0.8.6.2

Improved flexibility of prediction methods: "new" data frames do not need to contain exactly the same random effects identifiers as those used for model fitting, they may also contain a subset (as is useful for predmarg() from the "mpred" package.