Optimize inefficient database query in the deviation list view

The course module name is no longer queried separately for
each deviation. This reduces the number of performed
queries significantly. Additionally, some large unnecessary
attributes are no longer loaded.

Fixes apluslms#1302

deviations/viewbase.py

@@ -257,13 +257,18 @@ def get_deviation_groups(
             'submitter', 'submitter__user',
             'granter', 'granter__user',
             'exercise', 'exercise__course_module',
+            'exercise__course_module__course_instance',
         )
         # parent is prefetched because there may be multiple ancestors, and
         # they are needed for building the deviation's URL.
+        .defer(
+            'exercise__exercise_info',
+            'exercise__description',
+            'exercise__course_module__course_instance__description',
+        )
         .prefetch_related('exercise__parent')
         .order_by('submitter', 'exercise__course_module')
     )
-
     deviation_groups = groupby(
         ordered_deviations,
         lambda obj: (obj.submitter, obj.exercise.course_module),