Add a new pydough.explain_llm() function that returns a structured description of a PyDough collection expression, designed for use in LLM-assisted text-to-PyDough pipelines.

The existing explain() API produces human-readable description which can be difficult for a model to parse programmatically.
explain_llm returns a deterministic, machine-parseable output that an LLM can use to verify whether generated PyDough code correctly answers a user's question.