Navigating the Black Box: Leveraging LLMs for Effective Text-Level Graph Injection Attacks