Replace OpenAI proxy with LiteLLM

[ericmichael]

Summary

• add LiteLLM proxy configuration, docker wiring, and updated requirements plus a stateless config file
• teach the Agent and Thread flows to honor the LiteLLM model list, service-key fallback, and refreshed chat/message VCR tapes
• document the integration progress, new environment variables, and LiteLLM test workflow while dropping obsolete unit tests/cassettes

Testing

• DJANGO_SETTINGS_MODULE=aistarterkit.settings python manage.py test chat.tests.integration

Fixes #29