Multicalibration for LLM-based Code Generation