Response generation extremely slow

[madiator]

I see a for loop being used to generate the responses. This can be extremely slow.
For example, for the npc type, each response takes about 30 seconds with gpt-4o and so the whole thing on 200k personas can take about 69 days.
I have reimplemented an optimized version using bespokelabs-curator.
Please see here: https://github.com/bespokelabsai/curator/tree/dev/examples/persona-hub
If someone is on tier 5 with openai, the end result can be obtained in about 7 hours.
Happy to do a PR that uses this version.