Add aggregate_data method to ease pre-processing of data for bmmodels based on the bmmformula to analyze the data.

[GidonFrischkorn]

When working on implementing the m3 into the package, I thought about adding a aggregate_data S3 method that could ease pre-processing for bmmodels that can be estimated with aggregated data in each of the cells of the design. This is relevant to models such as the m3 or sdt models that use a binomial or multinomial distribution.

Currently, these models have not yet been added to the package, but with the package gradually growing I wanted to ask you @venpopov what you think about this?

For the final use I thought about something along the lines of:

agg_data <- aggregate_data(bmmodel, bmmformula, raw_data, ...)

For models, that have no aggregate_data method, we could just print a message, that these models are estimated on trial level data and thus cannot use aggregated data, and for the other models the function would aggregate the data based on the predictors used in the bmmformula. Potentially, some additional information would have to be added in ... to specify which variable contains relevant information in the raw_data.