Chain-of-Thought Reasoning in Streaming Full-Duplex End-to-End Spoken Dialogue Systems