Self-AMPLIFY: Improving Small Language Models with Self Post Hoc Explanations