For example, user may want to query all the runs for a prompt and see the llm_output and compare them.
Right now, they have to get the prompt, loop its versions, for each version, loop its runs, and the add the run to a list.
Ideally, they should call one method and get everything in a dataframe