Handling objective function

[cprudhom]

The current way of managing objective function is quite unsatisfying from modeling side and, in lesser extent, from solving side.
The main drawback is that it was designed to deal with single objective constraint problem but now multi-objective functions can be defined. Actually, there are two specific such functions that are problematic: Lexico and Pareto.

Here is a list of the current drawbacks (certainly not exhaustive):

1. The model stores a variable as the one to optimize; considering two or more variables is not possible
2. The parallel portfolio cannot be used with multi-objective functions
3. Default search strategies detect the objective in mono-objective case; the ones in multi-objective cases are ignored
4. When dealing with lexico or pareto, the COP is first turned into a CSP
5. IMeasures is designed to store the best value of a COP; storing and returning multiple objectives is not possible (there are other places, I suppose less visible, like VerbositySolving, in the same situation)
6. Even though an objective has been defined in a model (with model.setObjective(_, _), the dedicated method to find the optimal solution requires the objective and the policy to be defined again (this maybe out the scope though...)

I can live with that, but it can also be improved.