[Feature Request]: different factor levels for different split contexts

Feature description

similar to https://github.com/insightsengineering/chevron/pull/343 where different precisions are needed, we have similar cases that, when splitting by "parameters", the levels need to be adjusted.

library(dplyr)

data <- expand.grid(
  USUBJID = c("1", "2", "3"),
  ARM = c("arm a", "arm b"),
  PARAMCD = c("test a", "test b", "test c"),
  AVISIT = c("VISIT 1", "VISIT 2", "VISIT 3")
) %>%
mutate(
  AVALCAT = case_when(
    PARAMCD == "test a" ~ sample(c("high", "low"), size = n(), replace = TRUE),
    PARAMCD == "test b" ~ sample(c("normal", "abnormal"), size = n(), replace = TRUE),
    PARAMCD == "test c" ~ sample(c("high", "middle", "low"), size = n(), replace = TRUE),
    TRUE ~ sample(c("high", "low"), size = n(), replace = TRUE),
  )
) %>%
df_explicit_na

basic_table() %>%
  split_cols_by("ARM") %>%
  split_rows_by("PARAMCD") %>%
  split_rows_by("AVISIT") %>%
  summarize_vars("AVALCAT") %>%
  build_table(data)

even if we there are no actual categories in that parameter, the levels are there; pruning is removing all zeros so maybe not optimal if we want to keep some 0 count rows. Best if it could be achieved in analysis part.

Looping @barnett11 for your comments. This arise up when I review a template of egt05_qtcat_1 where Tim asks if by parameter analysis is possible

Code of Conduct

• I agree to follow this project's Code of Conduct.

Contribution Guidelines

• I agree to follow this project's Contribution Guidelines.

Security Policy

• I agree to follow this project's Security Policy.

Start with the linked chevron pull request 343, then run the example using basic_table(), split_cols_by(), split_rows_by(), summarize_vars(), and build_table(). Define how factor levels should vary by PARAMCD while retaining selected zero-count rows, and verify the resulting table against the demonstrated data and egt05_qtcat_1 use case.