estimation plots only show per database, when Bayesian Likelihood Approximation is used

[mvankessel-EMC]

This reprex pattern for CohortMethod, and this regex pattern for SelfControlledCaseSeries, make it so only Bayesian likelihood Approximation (i.e. from createBayesianMetaAnalysis()) is supported. If any other random approximation method is used, or no meta analysis is performed at all, it will overwrite data$database with NAs. This causes the results for different databases to stack on top of each other.