Umap behaviour is sklearn pipeline

Does adding umap in a sklearn pipeline which is finally called e.g. using `pipeline.fit(X,y)` produce a map in an unsupervised or in a supervised way? Is there a way to control for that? Thanks

Start by reviewing UMAP's scikit-learn transformer documentation and the behavior of sklearn.pipeline when fit receives y. Determine whether pipeline.fit(X, y) selects supervised or unsupervised behavior and how that choice can be controlled; done when the behavior and control mechanism are clearly documented.

Written by the indexing model from the issue text.