Reasoning on a Spectrum: Aligning LLMs to System 1 and System 2 Thinking