This is a follow-up to #31322 that added a classical MDS implementation as ClassicalMDS class. As discussed over there, this PR does the following:

1. It adds support for non-Euclidean metrics to the MDS class, following example of ClassicalMDS.
2. It changes the API of MDS to make it consistent with ClassicalMDS and other sklearn classes: distance metric is now set via the metric argument, and metric/non-metric MDS can be toggled via metric_mds=True/False argument. Backwards compatibility is ensured for the next scikit-learn versions.
3. Classical MDS is added as an initialization option to the MDS class, to be made default in the future.

PS. Apologies for lots of obsolete commits here :-( Just ignore them.