ViBe: A Text-to-Video Benchmark for Evaluating Hallucination in Large Multimodal Models