Llama-based source code vulnerability detection: Prompt engineering vs Fine tuning