Fix issue #72 - #74

Should there be a test which expects the proper error? Similar to test_set_invalid_method in test_timesolver.py

Hmm, perhaps, although I think this might be overkill (it's a delicate interplay between magnets in World and its associated TimeSolver). Not every std::runtime_error is caught in a pytest test. However, if you want to, a timestep of zero could be added to test_set_invalid_timepoints in test_timesolver.py. Or, which might be even better, define a new function test_invalid_timestep which catches even more invalid values.