Make analysis output more flexible (=individually configurable)

Right now, we have exactly *one* analysis interval in the `AnalysisCallback`, which controls both the calculation of the mandatory L2/Linf errors and the output to the files with some analysis quantities. As discussed back when we implemented this for the Jeans instability, we knew that this is not a sufficiently scalable approach for larger-scale calculations.

I have now encountered this limitation for 3D LBM calculations for the Taylor-Green vortex, where I need to write out integral quantities such as kinetic energy or enstrophy in regular time intervals. However, due to the way the current system is set up, this also means that each output interval time (which is every couple of time steps) I have to do a full analysis step - including the calculation of L2/Linf errors - which is very expensive for 27 conservative variables...

It is not an urgent matter, but we should think about what kind of capabilities we want from such a feature and then implement it properly.

Start by reading the AnalysisCallback and its current analysis-interval handling. Define the desired separation between mandatory L2/Linf error calculations and analysis-quantity output, then verify that integral quantities such as kinetic energy and enstrophy can be written at their own intervals without triggering the expensive full analysis each time.

Written by the indexing model from the issue text.