Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs