It takes 2 dataframes and a list of string corresponding to the primary keys and compute 3 dataframes: the new rows, the deleted rows and the updated rows