Overcoming Pitfalls in Graph Contrastive Learning Evaluation: Toward Comprehensive Benchmarks