Improve constraint violation visualization

[tjstienstra]

Fixes #123 and #53

Changes

Correct the slicing of the constraints_violation vector.
The constraint vector is formatted as (refer to the creation of the constraint function and the wrapper:

[con_1_2, ..., con_1_N,
con_2_2, ..., con_2_N,
...,
con_n_2, ..., con_n_N,
*instance_constraints]

where con_1_2, ..., con_1_N the constraint violation is at every node for the first discrete equation of motion.

Only create a single axis for showing the EoM violation, as they are not officially coupled to specific states. Yes, it is correct to say that most will specify the first n//2 as kinematic differential equations, but that is not required.

Only plot the constraint violations of the instance constraints if there are any instance constraints.

[tjstienstra]

Here is the new version of the pendulum from the examples page

And here are the plot constraints of the first optimization problem in the BRiM paper

[moorepants]

Is setting the x tick labels at 90 degree rotation better, in general? If so, maybe adjust that while you are at it.

[tjstienstra]

Is setting the x tick labels at 90 degree rotation better, in general? If so, maybe adjust that while you are at it.

I don't like those ticks in general, they always make a mess especially if a constraint is a longer equation. In this case, if you would choose -90, you actually run into the problem that a part of your equation falls outside (below) the figure.

[tjstienstra]

It is that you like to see which one it approximately is, but otherwise I would just throw them all together in a single boxplot

[moorepants]

We can go with this as is and address formatting later.

[moorepants]

Forgot about this. Merging.

[moorepants]

The docs show this for the pendulum swing up:

which is quite smooth compared to the old:

[moorepants]

Also isn't as small.