fix: guard choices[0] and message=None before content access

OpenAI-compatible APIs can return empty choices on error or content
filtering; Gemini returns choices[0].message=None on PROHIBITED_CONTENT
(HTTP 200). Both cause unhandled exceptions at runtime.

Example files: adds `if not response.choices or response.choices[0].message is None: raise ValueError(...)`.
Test files: adds `if not ... : self.fail(...)` guard before each assertion.