#226 Fix acceptance tests

[Ulf Lorenz]

What and Why

On my new machine, some of the acceptance tests fail. These should be fixed and/or the tests generally be made more robust.

• Plotting output of eigenstate calculation produces th same state with negative sign
• In some tests there are differences in the wavefunction: "-0.00000" instead of "0.00000".
• minute differences in the plotting output on the order of 1e-6

Also, the wording in the various Readmes and such should be changed to ensure that an acceptance test failure is not a catastropic error, but may happen depending on the machine.

---

• mostly disabled the offending output completely. It is not terribly useful anyway (1D only, Python interface is way more flexible here)
• added dedicated demo tests for the output with well-defined states that are plotted or output.
• added ability to mark certain output artefacts as "binary", in which case only their existence is checked, not the hash.
  That means, though, that you need to check these manually, for example before a release.