Automatically check if optimisation succeeds or fails (infeasible models)

[p-snft]

It would be easier to pin down issues like #1045 if we checked the status of the optimisation before collecting the results.

More context:
When an energy system model is infeasible, solph.Model.solve() will just warn ("UserWarning: Optimization ended with status warning and termination condition infeasible") return without an error. Then, latter solph.processing.results(model) will just seem to work (but not return meaningful results), but processing.meta_results(model) throws a KeyError. This is because it cannot find the objective value in an infeasible model.

As you see, the error handling is not intuitive at all. I see some options:

Model.solve():

• Warn if the problem is infeasible.
• Fail if the problem is infeasible. (I would opt for this, as solving fails.)

processing.results() with (non-existing) "results":

• Warn if there are no results. (I would opt for this, as the function can be used to extract input data. Although it is not the design way to do things, I don't see why we should disallow this. If solving already fails, it is unlikely that somebody does this existentially. Thus, a warning should do.)
• Fail if there are no results.

As it will fail anyway, there are no options for processing.meta_results(). We should handle that in our code and provide a meaningful error message.

[p-snft]

Might be a topic for oemof/oemof#124