Added tests to verify that the stack:
1. Continues initialization when providers fail to register models
2. Skips invalid models instead of crashing
3. Handles provider listing failures gracefully
4. Maintains partial functionality with mixed success/failure
Example:
- OpenAI provider fails to list models
- Stack logs error and continues with registered models
- Other providers remain functional
This prevents the entire stack from crashing when:
- Provider API keys are invalid
- Models are misconfigured
- Provider API is temporarily unavailable
