Lagrange objective should be calculated at each integration steps

[pariterre]

Currently the objective Lagrange functions (and constraint for MINIMIZE) are only computed at the nodes (as for the continuity constraints. However, this approximation may gets ugly if the function is highly nonlinear and the shooting nodes are spread apart.

A solution would be to use the intermediate points of the integration (for ERK at least).

[pariterre]

Also this integration should be optional (the current behavior should be available) and the type of integration should be accessible (Piecewise constant, Trapezoïdal, etc.)

[pariterre]

@Ipuch Check what happens in COLLOCATION

[Ipuch]

It's the same with COLLOCATION. As an intermediate way, one could use #458, trapezoidal integration instead of rectangle integration.