Feature Request: Partial Least Squares Discriminant Analysis macro

Sklearn implements Partial Least Squares Regression (PLSR) but a very common use for this algorithm is a slight variation called Partial Least Squares Discriminant Analysis (PLS-DA), which is when y is categorical instead of continuous. You can pretty easily write PLS-DA using the PLSR from sklearn, as documented in the first answer to this question but it seems silly that everyone who would want to do PLS-DA has to implement this. Other libraries in R, Matlab, etc... provide an interface to PLS-DA directly, and I think sklearn should too. As I imagine it, this would be a very simple method which just does some LabelEncoding and then calls PLSR, but it would save a lot of folks like myself a lot of time.

Start with sklearn.cross_decomposition.PLSRegression and the linked Stack Overflow implementation to understand the proposed PLS-DA interface. Clarify how categorical y should be encoded and what tests and documentation are required; the work is done when scikit-learn provides a defined PLS-DA API with verified categorical-target behavior.