[feature request] grouped overall columns

the user would like to be able to do something like:

```r

adsl_mod <- syn_data$adsl %>%
mutate(collapse_TRT = case_when(TRT01P %in% c("A: Drug X")~ "A",
TRT01P %in% c("B: Placebo", "C: Combination")~ "BC"))

result <- basic_table(show_colcounts = TRUE) %>%
# Split columns by treatment arms
split_cols_by(var = "TRT01P") %>%
# Add custom total columns for selected treatments
add_overall_col("Total by treatment grouping") %>%
split_cols_by(var = "collapse_TRT") %>%
add_overall_col("All Patients") %>%
# Analyze variables
analyze_vars(
vars = c("AGE", "SEX", "ETHNIC", "RACE", "BMRKR1"),
var_labels = c(
"Age (yr)",
"Sex",
"Ethnicity",
"Race",
"Continous Level Biomarker 1"
)
) %>%
build_table(adsl_mod)

result
```

Start by tracing the existing basic_table, split_cols_by, add_overall_col, analyze_vars, and build_table entry points to understand how column groups are represented. Compare the example's grouped totals with current behavior and define tests showing that selected treatment groups and the all-patients total are produced correctly.

Written by the indexing model from the issue text.