The disk.frame group by operation perform group WITHIN each chunk. This is often used for performance reasons. If the user wishes to perform group-by, they may choose to use the `hard_group_by` function which is expensive as it reorganizes the chunks by the shard key.
# S3 method for disk.frame group_by(.data, ...) chunk_group_by(.data, ...)
same as the dplyr::group_by