pick first matching question when creating previous links

In theory this should not happen but this is to resolve some data issues
with initial scorecards.

crowdsourcer/management/commands/link_questions.py

@@ -69,8 +69,9 @@ def get_questions(self, session):
         questions = defaultdict(dict)
 
         for section in Section.objects.filter(marking_session=session):
-            for question in Question.objects.filter(section=section):
-                questions[section.title][question.number_and_part] = question
+            for question in Question.objects.filter(section=section).order_by("pk"):
+                if questions[section.title].get(question.number_and_part) is None:
+                    questions[section.title][question.number_and_part] = question
 
         return questions