Function Calling - Revised Reduce Operation

[nikhil23-Rao]

• changing the output schema of reduce operation in folded outputs (to ultimately reduce # tokens llm has to output)
• edited cached_call_llm function which has logic for processing function calls and storing it in an updated state
• at the very end of all the folds convert the localized state into the user's self-defined schema in the reduce.py file
• testing data all moved to a folder called testing-data to avoid pollution of project in base
• found certain experiments with the code to result in faster times and cheaper costs