Document equivalence of re_formula = NA and re_formula = ~0 in predict methods

[wpetry]

Users can specify which group-level "random" effects to condition on when making predictions. brms is consistent with other mixed modeling packages (lme4, rstanarm, etc.) in allowing (i) a one-sided formula specifying the effects to condition on, (ii) NULL to condition on all effects in the model, or (iii) NA to include no group-level effects.

Like the other packages, brms accepts re_formula = ~0 as an equivalent to re_formula = NA, but this is not documented for the predict.brmsfit method.