Transforms wide enrollment data to long format with subgroup column.
Value
A long data frame with `value_source` describing count provenance. Published counts use `"published"`. Where a workbook supplies only a rate, the count uses `"derived_from_pct"` when the same row's published total is available and `"published_pct_only"` otherwise. The `free_reduced_lunch` subgroup inherits the weaker of its component labels. From 2023, published program counts are available by grade for the state and at all grades for districts and schools. `published_pct` contains a workbook's published program rate or statewide Non-Binary rate as a fraction where available. `pct` is the count divided by the row total. The 2024 statewide `non_binary` rows carry `value_source = "not_reported"` with `n_students`, `pct` and `published_pct` all `NA`: NJ DOE's 2023-24 file printed 0 in every Non-Binary cell, a category it did not collect that year.