[Dev] Add parameters #655

FTR:

• for indirect matters, variable should be enough @gergaud
• for direct it is a priori not and there is a simple way to define parametric optimisation models, e.g., in ExaModels.jl (check this - no sensitivities, though)
• for direct, returning a function (of any parameter one wants) towards a model is easily done and may be not too harmful for performance; unless one wants more (sensitivities through suited AD)
• if necessary: extend the syntax by allowing

    @def begin
        p in R^r, parameter
        ...
    end