[REFACTOR] Hardcoded OpenAI Models

[ebhills]

What needs refactoring
calls to specific OpenAI models in our code and tests

Current state
hard coded to older models. Some will probably break soon b/c of deprecation. Also, creates lots of noise in our usage and cost data.

Proposed changes
Make models a top-level, global CONFIG VAR. Point all AI functions and tests at this VAR.
Create specific keys for various functions (eg PY Testing, Schema Standardization, ...) so that we have granular visibility of usage / costs.

Benefits

• Performance improvement
• Better maintainability
• Code quality
• Other:

Breaking changes

• Yes
• No

Details

Acceptance criteria

• [ ]
• [ ]
• [ ]