The mark_part JME function copes with pre-submit

Gap-fills run `mark_part` on all of their parts. We can't know if
they're going to have pre-submit tasks until after trying to mark them.

This changes the `mark_part` function to return an empty result if the
part is waiting for pre-submit tasks. The gap-fill's own pre-submit note
then waits for all the gaps' pre-submit tasks to finish before trying to
mark again.

Author: christianp

runtime/scripts/marking.js

@@ -408,6 +408,17 @@ Numbas.queueScript('marking',['util', 'jme','localisation','jme-variables','math
             } else {
                 var part_result = part.mark_answer(answer, part.getScope());
             }
+            if(part_result.waiting_for_pre_submit) {
+                return jme.wrapValue({
+                    marks: part.availableMarks(),
+                    credit: 0,
+                    feedback: [],
+                    valid: false,
+                    states: {},
+                    state_valid: {},
+                    values: {}
+                });
+            }
             var result = marking.finalise_state(part_result.states.mark);
             return jme.wrapValue({
                 marks: part.availableMarks(),