Weak-eval-Strong: Evaluating and Eliciting Lateral Thinking of LLMs with Situation Puzzles