Create `aggregate_data` function based on `bmmodel` and `bmmformula`

[GidonFrischkorn]

This function should take the bmmodel and the bmmformula and then aggregate the data to optimize model estimation. This will speed up model estimation for all bmmodel that do not need to fit trial wise data

[chenyu-psy]

An idea of the function:

tidy_M3 <- function(
    .data,
    responses,
    group = NULL,
) {

  # Add response column to group
  new_group = c(group, "response")

  # Aggregate responses
  data_output <- .data %>% 
    select(all_of(responses), all_of(group)) %>% 
    pivot_longer(cols = all_of(responses), names_to = "response", values_to = "value") %>% 
    summarise(
      Resp = sum(value, na.rm = TRUE),
      .by = all_of(new_group)) %>% 
    mutate(
      {{nDV_name}} := sum(Resp, na.rm = TRUE),
      .by = all_of(group)) %>% 
    pivot_wider(names_from = response, values_from = Resp)

  return(data_output)

}

[chenyu-psy]

This function should be applicable to all multinomial distribution data: https://github.com/chenyu-psy/smartr/blob/main/R/agg_multinomial.R

[GidonFrischkorn]

For now, I have decided not to implement this only for the m3, but I have created a discussion in the bmm repository: #236