CI: LLM acceptance openai + litellmrouter jobs fail — gpt-4.1-nano retired by OpenAI #6101

Since ~2026-10-25 the LLM acceptance (openai/gpt-4.1-nano) and LLM acceptance (litellmrouter/gpt-4.1-nano) jobs fail on every PR that touches the API (e.g. run 40298512661 on #6100, and the fix/maintenance-stop-promptly run).

What happens: every call returns HTTP 404: The model gpt-4.1-nano has been deprecated, so test_llm_api_methods, test_retain_with_entity_labels_under_strict_schema, test_real_llm_retain_and_consolidation_traced and test_reflect_searches_mental_models_when_available fail.

Should happen: the matrix runs against a live OpenAI model.

Where: .github/workflows/test.yml LLM acceptance matrix (the two model: gpt-4.1-nano entries). Swap to a current small OpenAI model, and check docs/defaults for other references to gpt-4.1-nano.