What's the feature?

Right now if I have two ARDs one with a list column and one not I need to do some manual work to bind them; both bind_rows and bind_ard will error in this case. See this artificial example:

library(cards)

ard_auto <- ADSL |>
  dplyr::group_by(ARM) |>
  ard_summary(
    variables = "AGE",
    statistic =
      ~ list(conf.int = \(x) t.test(x)[["conf.int"]] |>
               as.list() |>
               setNames(c("conf.low", "conf.high")))
  )

ard_custom <- ADSL %>% 
  summarise(
    stat = median(AGE)
  ) %>% mutate(
    variable = "AGE",
    stat_label = "Median",
    stat_name = stat_label
  ) %>% as_card()

cards::bind_ard(ard_auto, ard_custom)

ard_custom_list <- ard_custom %>% 
  mutate(
    stat = list(stat)
  )

cards::bind_ard(ard_auto, ard_custom_list)

I think that either

1. as_card() should either be more opionated (reject columns with wrong format) or transform columns
2. bind_ard() can handle if it sees a list column and a numeric column

To some extent I feel like bind_rows should really be able to handle this, but seeing as this is likely to come up I think a handler.

My personal preference is for as_card() to be a bit more demanding to help me understand whether the tibble I have made is actually close to being a fit for cards.