Fix test_quickcomplete_flicker

The test needed to be fixed because of how the completer behaviour
changed.

Before:
completer always scheduled a completion update on selection changed,
but the updates themselves were ignored if not needed.

Now:
completer only schedules completion updates when actually needed, but
never ignores a completion update.

So, before it was correct to check whether `set_model` was called, now
we must check if the completion was actually scheduled. This can be
done by checking the parameters with which `_change_completed_part`
is called, since a completion is scheduled only when `immediate=True`

tests/unit/completion/test_completer.py

@@ -299,6 +299,12 @@ def test_quickcomplete_flicker(status_command_stub, completer_obj,
     config_stub.val.completion.quick = True
 
     _set_cmd_prompt(status_command_stub, ':open |')
-    completer_obj.on_selection_changed('http://example.com')
-    completer_obj.schedule_completion_update()
-    assert not completion_widget_stub.set_model.called
+
+    url = 'http://example.com'
+    completer_obj._change_completed_part = unittest.mock.Mock()
+    completer_obj.on_selection_changed(url)
+
+    # no immediate (default is false)
+    completer_obj._change_completed_part.assert_called_with(url,      # text
+                                                            ['open'], # before
+                                                            [])       # after