Pseudocode-Injection Magic: Enabling LLMs to Tackle Graph Computational Tasks