Port from pydantic-ai to openai-agents SDK

[pamelafox]

Purpose

I wasn't able to fix pydantic-AI to remove the extra call after the tool call, so I ported to the openai-agents SDK, which has built-in support for that via the tool_use_behavior="stop_on_first_tool", argument to the Agent. This seems to be working fairly well, though I've seen that llama3.1 from Ollama is only succeeding half the time (sometimes it returns JSON in the content string instead of as a tool call). Not sure what's affecting the reliability of the Ollama SLM calls.

Does this introduce a breaking change?

When developers merge from main and run the server, azd up, or azd deploy, will this produce an error?
If you're not sure, try it out on an old environment.

[ ] Yes
[X] No

Type of change

[ ] Bugfix
[X] Feature
[ ] Code style update (formatting, local variables)
[ ] Refactoring (no functional changes, no api changes)
[ ] Documentation content changes
[ ] Other... Please describe:

Code quality checklist

See CONTRIBUTING.md for more details.

• The current tests all pass (python -m pytest).
• I added tests that prove my fix is effective or that my feature works
• I ran python -m pytest --cov to verify 100% coverage of added lines
• I ran python -m mypy to check for type errors
• I either used the pre-commit hooks or ran ruff manually on my code.