Implemented interface to FieldsIO in core problem class and respective hook

[brownbaerchen]

I added three functions to the core problem class, which I hope cover most cases:

• setUpFieldsIO: Mostly for setting up space decomposition. Here, you can call all the class methods of FieldsIO and derived classes
• getOutputFile: This function is supposed to instantiate a FieldsIO object fully set up, only taking a file name as an argument. This needs to be done individually for the problem classes to set the header and data structure correctly
• processSolutionForOutput: Do stuff to the solution before writing to the file. For instance, when using spectral methods, you may want to transform from spectral space to real space and throw away the imaginary part to save disc space.

Once these functions are set up for the problem class you want, you can simply add the hook LogToFile (I renamed the previous hook of this name to LogToPickleFile).
You configure this class via class attributes for file name, simulation time increment between writes and whether or not you allow overwriting of previous files.
This hook also supports restarting runs. If the path to the file exists and you are starting from a time greater than zero, it will append the file with the new solutions.

I added the setup of FieldsIO for generic_spectral and TestEquation0D classes. I have used this to run the 3D RBC examples that I will merge at some point.

Finally, I wrote a bunch of tests for this. Note that I am using pytest-isolate-mpi as discussed in #539 here.

This PR works on #534. It is addressing functionality parts rather than documentation, which should be added later as well.

[brownbaerchen]

Does anyone have any opinions on this? :)

[tlunet]

I'm fine with merging it. especially since @brownbaerchen started to use more the camelCase convention 😍