Python Pandas drop row duplicates on a column if no duplicate on other column