Fix the total returned by grade_gateway when the best score is zero.

[drgrice1]

Previously a set version only replaced the initial [0, 0] best set data if its score was greater than zero. So if a user scored zero on all versions of a test, or had not started the test, the total returned was also zero. For LTI 1.3 grade passback that resulted in a scoreMaximum of zero being sent, which Canvas rejects with "cannot be zero if line item's maximum is not zero".

Now the first version is used if no version has a positive score, and if there are no versions the total is computed from the problems assigned to the user for the test.

Aside: I have been seeing the message "Failed to send grade: Unprocessable Entity" quite frequently because of this. Changing line 366 of lib/WeBWorK/Authen/LTIAdvantage/SubmitGrade.pm to $self->warning(join("\n", 'Failed to send grade:', $response->body)); gives the more meaningful message above. Note this message issue I am looking for a better solution for. The real problem is Moodle doing things quite badly. It seems that all other LMS's give a JSON response. Canvas literally gives {"errors":{"type":"unprocessable_entity","message":"cannot be zero if line item's maximum is not zero"}} in this case. I am rather certain that D2L also gives this sort of JSON response from what I have seen when @dlglin and @Alex-Jordan test things. Moodle on the other hand gives an HTML response that has a body that is quite large and useless, and the $response->message is a brief and more meaningful message. Perhaps this should just be switched to $response->body and make Moodle users decipher the big HTML content (the message is hidden in it I believe)? Although, maybe Moodle has fixed this. I don't think I have tested this for a while with Moodle.

[drgrice1]

I tested with Moodle now, and even Moodle sends a JSON response with a more informative message using $response->body. Either that has changed or Moodle alway did this for this request, and I was confusing this with another LTI request. In any case, I added that to this pull request.