How Good are LLMs at Relation Extraction under Low-Resource Scenario? Comprehensive Evaluation