Add a guardrails system to nilAI

[jcabrero]

This involves adding Llama Firewall as an option to nilAI queries. This will be a specific guardrail feature that users will be able to enable to classify their query based on maliciousness.

http://github.com/meta-llama/PurpleLlama