How to avoid duplicate columns after join in PySpark