epi_archive's compactify doesn't support distributions

[dsweber2]

mwe:

tibble(ahead = 2L, 
  geo_value = "ak", 
  forecast_date = as.Date("2020-01-17"), 
  target_end_date = as.Date("2020-01-19"), 
  actual = 25, 
  .pred_distn = epipredict::dist_quantiles(list(c(1,5,9)), c(0.1,0.5,0.9))) %>% 
 as_epi_archive(other_keys = "ahead", time_value = forecast_date, version = target_end_date)

Unclear if we want to support this and/or weirder column types in general. Logan was suggesting solving via vctrs.