Update nonlinear channel flow benchmark

[gassmoeller]

This is a follow-up PR that is based on #6159. I here also rework the nonlinear channel flow benchmark. I still need to check some tests, but one thing I already noticed is: If I change the boundary traction value or the boundary velocity value the solver converge much better than with the values they are set to know. I kept the values the same to reproduce the figure like they are in the original paper, but I wanted to record this somewhere. It is curious how the traction benchmark stagnates at a constant nonlinear solver tolerance, even for models that use the standard Picard solver (therefore it is not related to the Newton solver or the defect correction solver).

[gassmoeller]

Ok, @MFraters I think this is also ready for a review. This PR does the following things:

• Simplify running the nonlinear_channel_flow benchmark.
• Expand the documentation of the benchmark, in particular mentioning the difference between the figures in the original Newton solver paper and what the current default solver parameters do.
• Add 2 tests that use exactly the benchmark input files. This will ensure the benchmark files stay usable, and will also alert us if the results change. Although we already have many tests for the nonlinear channel flow benchmark they use their own copy of the input file and are therefore way out of date. I thought about changing the tests to use the existing benchmark file, but they are all based on one main test file, and because of Support recursively including parameter files in a better way #6165 I cannot simply base the main test file on the benchmark files.