* Reset the global registry after each test (teardown)
* Create a settings fixtures that returns graphene_settings and resets
the graphene_settings after use (teardown)
* Convert test_mutation tests from unittests.TestCase to pytest
* Convert test_mutation PetType to a pet_type fixtures that reregisters
the type
When we use the `graphql_schema` management command, the output can vary from run to run depending on arbitrary factors (because there is no guarantee made about the order used to output JSON dictionary keys). This makes it difficult to compare two schema's at different points in time.
We address this by including a new `canonical` flag to the command, which uses standard `json.dump` funcitonality to sort dictionary keys and force pretty-printed output.
