prevent (or add option to prevent) stratum/lode positions from depending on missing value inclusion/exclusion

[corybrunson]

Currently, setting na.rm = FALSE affects the vertical placement of strata. Ideally, it would not, so that, for example, the stratum layer could omit entries with missing values and the alluvium layer retain them. See the example below, which illustrates the incompatibility.

# date-valued axis variables
vaccinations |> 
  transform(response = dplyr::if_else(response == "Missing", NA, response)) |> 
ggplot(aes(x = end_date, y = freq, stratum = response, alluvium = subject,
           fill = response)) +
  stat_alluvium(geom = "flow", lode.guidance = "forward",
                width = 60, na.rm = FALSE) +
  stat_stratum(width = 60, na.rm = TRUE) +
  labs(x = "Survey date", y = "Number of respondents")
#> Error in ggplot(transform(vaccinations, response = dplyr::if_else(response == : could not find function "ggplot"

^{Created on 2025-08-16 with reprex v2.1.1}

Note: This behavior arose while experimenting with this SO question.