Next opty

[Peter230655]

I feel, based on my playing around, that opty has taken quite a leap by

• allowing unknown input trajectories to have bounds
• allowing equations of motion to have bounds
• allowing more eoms than state variables

Unknown input trajectories cannot be differentiated, but this may be a blessing in disguise:
When bounds became available for input trajectories and I replaced state variables by input trajectories execution was always faster (I reported this before, #449). I have no idea why, just an observation on a number of simulations. If opty gets changed so that input trajectories may be differentiated, maybe they become slower. I see no need to take this risk.