PySpark: How To Call geocoding api Without Hitting 50 QPS limit with maximum parallelism possible?

Question

Currently I have modified geocoding daily limit to unlimited per day as i have to process existing 5 million addresses as quickly as possible.

I am use aws glue job t