SPOR step checks if observations are present

[clemens-fricke]

Is your feature request related to a problem? Please describe.
Currently, the SPOR steps do not check if the observations defined for them are actually produced in each step. This will make the training crash with weird error messages.

Describe the solution you'd like
two possible solutions are possible

1. Define default observations if no/too few observations are provided. (What do we do if there are too many?) AND write a warning to the log saying that observations were amended with default observations. The episode should also be terminated at this point.
2. Raise an error saying this specific step did not provide the observations that it should.

This is currently seen as a nice to have but not an urgent feature.