Can Large Language Models Identify Implicit Suicidal Ideation? An Empirical Evaluation