Fix a possible ordering bug in label test

Instead of adding a key to a set (which has no order), and later testing
against that order, just use a list for deterministic tests.

ws/tests/views/test_trips.py

@@ -439,7 +439,7 @@ def test_updates_on_stale_trips(self):
         self.assertIn(
             'To make updates to the trip, please load the page again.', warning
         )
-        self.assertIn('Fields which differ: Description, Leaders', warning)
+        self.assertIn('Fields which differ: Leaders, Description', warning)
 
         # No edit was made; we have form errors
         trip.refresh_from_db()

ws/views/trips.py

@@ -368,15 +368,15 @@ def _stale_revision_message(self, form, current_trip, new_trip) -> Optional[str]
         if current_trip.edit_revision == new_trip.edit_revision:
             return None
 
-        fields_with_difference = {
+        fields_with_difference = [
             field
             for name, field in form.fields.items()
             if name != 'edit_revision'
             and getattr(current_trip, name) != getattr(new_trip, name)
-        }
+        ]
         # (Account for the fact that we might have stripped `leaders`)
         if 'leaders' in form.cleaned_data and self._leaders_changed(form):
-            fields_with_difference.add(form.fields['leaders'])
+            fields_with_difference.insert(0, form.fields['leaders'])
 
         if current_trip.last_updated_by is None:
             # This shouldn't ever happen, but the data model makes it possible