[BUG] Dealing with failed model fits

[TomDonoghue]

In the Python code, the model can fail to fit, leaving NaN results, and this happens cleanly (there is no error, and the code continues, etc). However, in the Matlab wrapper, it seems if a model fails to fit, the wrapper hits an error and fails. I forgot to check / test this case, so I think the wrapper just doesn't deal with the NaNs very well.

ToDo: check through how the wrapper deals with NaN parameters, and see if we can update it to work cleanly (continue) in this case.

Note: issue brough up by Matteo Fraschini