Probability of an LLM Trajectory: Starting from the Chain Rule