Deep Learning-Based Identification of Inconsistent Method Names: How Far Are We?