branching pipeline for ml demo notebook

currently there are three pipelines that are mostly identical. we could instead have a train-validate-test branching operator that returns three outputs from one pipeline, as currently done in hadisd notebook

Start by comparing the three currently similar pipelines with the hadisd notebook, which the issue identifies as the existing example. Trace how the proposed train-validate-test branching operator would connect to the ML demo notebook, and consider the work complete when one pipeline provides the three requested outputs instead of three mostly identical pipelines.

Written by the indexing model from the issue text.