Include some form of Dispersion tests in the testpackage

[ykempf]

Dispersion is very sensitive to anything done to the solvers, as the Vlasiator Summer of Debugging 2016 has shown.

To avoid having to put in the extra code, a version of the Dispersion tools suite using analysator would be very useful.

• [ ] Select suitable cases (quasi-parallel as done in August 2016, some quasi-perpendicular, some beam instability?).
• [ ] Develop a python-based FFT processing tool.
• [ ] Integrate in the testpackage, enabling both a visual and a numerical comparison.

General comment from @ykempf: beware of the file flood when doing the FFT, ensure they get removed after the test as they will likely be big.

[galfthan]

We could perhaps add a new function/hook into projects, that enable them to do some custom processing/IO during the simulation. In base class it would do nothing, but for dispersion it could write out the more efficient binary ones.

[ykempf]

Well that means an extra piece of code in the main propagation loop...

[ykempf]

I take Urs' comment on Flowdock as a volunteering to be assigned this issue.

[ursg]

I was just about to comment here. Indeed. :)

[galfthan]

It would be one extra line, so not too bad. My phiprof merge cleans out more lines than that ;) A project in itself would be to clean up main() to make it self explanatory and failry easy for somebody to understand