Support for handling censored data in PPC plots via imputation

[avehtari]

Currently, brms drops the censored data for most pp_check types. A better approach would be to impute the observations. See a bayesplot issue stan-dev/bayesplot#319 for a proof-of-concept and example plots. The imputation part would be natural to be in brms and bayesplot would just support a different color for imputed observations.

[avehtari]

Just added better figures to stan-dev/bayesplot#319

The imputation part requires sampling from a truncated posterior predictive distribution. In many cases rejection sampling is likely to be fast enough, and simplest to implement, but still requires some checks.

[hansvancalster]

This seems to be related to my proposal in #1657. With the data augmentation / imputation approach for modelling censored data, the censored data would be imputed during model fitting and could then probably be used in posterior predictive checks?