ENH use Xw - y instead of y - Xw as dual point

[mathurinm]

This is the classic Fenchel-Rockafellar primal dual link. It will be easier to check the maths, notably in the enet case.

[Badr-MOUFAD]

I think we better stick with the convention y-Xw. It is adopted entirely in code, plus all [my] the established equations follow it. So, there are no conflicts between what is implemented in the solver and the [new] equations.

P.S: refer to the maths snippets