Stochastic examples with trapezoidal transcription doesnt give exactly the expected values

[Ipuch]

Following:

• [RTR/RTM] Algebraic states are states #907
• [RTR after the others] moving the stochastic variables from a -> u #933

The examples of arm_reaching_muscle_driven.py, arm_reaching_muscle_driven_explicit.py, arm_reaching_muscle_driven_implicit.py won't return the values of the Leuven Implementation, because some of the stochastic variables are piecewise constant controls instead of the linear continuous across integration intervals.

This is due to a current limitation of bioptim that does not allow different types of controls for different decision variables (piecewise constant controls and linear continuous) for the same problem.