Towards Secure Intelligent O-RAN Architecture: Vulnerabilities, Threats and Promising Technical Solutions using LLMs