Modify the llms.py to make a goal satisfaction list function that is better suited for an intake questionnaire

[nonprofittechy]

Some things it needs to handle:

1. We will provide a list of protocols to the AI. these won't be visible to the user
2. The goal of the AI isn't the same as the expressed goal to the user (to say why they're there). The AI's goal is to answer if they qualify or not, and explain why.
3. We want to get a narrative response from the AI, not just a synthesized version of the user's statements.
4. We want to let the author provide the possible matching criteria in an unstructured form, not just as a list of individual goals. These aren't normally that long.

It does have a lot of similarities with the goal satisfaction list.

Related: maybe we just need an object that represents a back and forth chat, because the intake and satisfying a goal problems are similar but there could be even more kinds of conversations that don't easily fit these 2 structures.