36 revision of objective handling

[sarahleidolf]

• Introduction of subclasses for different sub-objectives for more intuitive parameterization of weighting factors
• Visualization of sub-objective results in a dashboard
• Original notation of the objective function still supported. “r_del_u” notation no longer supported.

[EserSteffen]

Hi, I finished the first round of review.
Generally really cool pull request, I like the overall structure and how the new examples work / look.

I've structured my feedback into points that should be addressed before this can be merged, and points that shouldn't block this merge, but may be considered in future issues.

Required now:

• remove all the AI-error handling and hasattr checks that catch errors from within the framework. Our error handling should target user errors. library errors should be raised so they are caught in CI.
• in the discretizations, extract duplicated code into functions, especially regarding the delta-u penalty. this can also be a lot more concise when the AI boilerplate / error handling is removed
• We can discuss on this, but I would prefer different names for the objectives, i.e. instead of EqObjective something like ExpressionObjective, or even only Objective, since this would be the standard case. But in general I like to avoid abbreviations in variablenames.

For the future:

• the current objective re-calculation is only exact for multiple shooting / euler, which should be acknowledged. In the future we might move this to the discretization, logging the points where objectives are evaluated and maybe tracking the exact calculation. E.g. for collocation its quite complex since we have weighting based on the collocation matrices.
• We now save 3 separate files for the mpc results. I think I would have preferred to keep the objectives with the stats. However we might go even further and move everything into a single file. This could be a separate issue and for now I wouldn't block the PR over this so its fine.
• We then - if saving objectives over the actual grid on a horizon - could even plot the predictions of the objective evaluation like variables in the dashboard.