Discretize using concrete operations with zonotopes

[mforets]

For the zonotopes set representation, Minkowski sum and linear map can be computed efficiently. The proposal is to offer a way to compute

Ω0 = ConvexHull(X0, ϕ*X0 ⊕ δ*U ⊕ □α)

in discretize returning the convex hull of two zonotopes. Then, we can use the existing function in LazySets.Approximations.overapproximate to convert it back to a zonotope.

[mforets]

This option applies to different approximation models, so i think that it is better to add it as a separate option. What woud be a good name? The option refers to the set type, or representation, used for the initial states and for the transformed inputs. Now we are using "lazy" set types: the initial states of the discretized system are a lazy convex hull.

[schillic]

We can add a new algorithm, for instance of the name "zonotopes".