Update default LLM model and improve fallback logic and error handling

[ritwikshanker]

This pull request updates the default large language model (LLM) used throughout the project from cesnet/qwen3-coder to cesnet/agentic and improves error handling for LLM provider failures. The changes ensure consistency across configuration, documentation, and test scripts, and make the fallback logic more robust by handling a wider range of provider-side errors.

Model Update and Consistency:

• Changed the default LLM model from cesnet/qwen3-coder to cesnet/agentic in the configuration, documentation, and all relevant code and test scripts. This ensures the new model is used as the primary LLM across the application and its supporting materials. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

Error Handling Improvements:

• Expanded the fallback mechanism for LLM calls to handle all provider-side errors (not just rate limits), by catching OpenAIError instead of only RateLimitError. This makes the system more resilient to various provider outages or misconfigurations. [1] [2]
• Updated documentation and code comments to clarify that the fallback model is now triggered on a broader set of errors, not just rate limiting. [1] [2]

These changes collectively improve reliability and ensure the application consistently uses the intended LLM model.