Randomly Dropping n-Groups from a Pandas DataFrame: A Correct Approach Using Series.unique and numpy.random.choice