Changed interaction logic for custom feedback and cases

[KarlLundengaard]

Previous behaviour was that if the evaluation function produced custom feedback this feedback was returned instead of the matched case feedback. The new behaviour is that the evaluation functions custom feedback is appended to the case feedback, unless the case has the parameter "override_eval_feedback" set to True in which case the evaluation functions custom feedback is ignored.

A few remarks:

• Corresponding documentation has not been updated to reflect the change
• The relation between custom feedback in the evaluation function and in the cases needs further work. In particular, if a case is matched when the evaluation function returns "is_true": True but the corresponding case might still be incorrect w.r.t. to the actual problem. Thus if we want to append the custom feedback from the evaluation function the custom feedback cannot say anything about whether the result is correct or incorrect, since this could lead to self-contradictory feedback.
• Some more information should probably be displayed to problem authors when running tests, at least what case is matched and what feedback will be shown.

[peterbjohnson]

This looks good to me so I'm approving it.

I'm concerned about the naming of 'custom feedback' - I think we need a name that has a more obvious meaning to someone who doesn't know what it is. I know this is not directly related to the pull request so we can discuss elsewhere.