Tests reference stale token formats

If you look through the current keystonemiddleware tests you'll see references to PKI and UUID token formats [0]. While these tests were written when those token formats were supported, it can be confusing to see references to them because it implies there is suppose somewhere for those things.

Instead of referencing implementation details of another service, we could try and generalize those properties and write the test data around those instead. This would prevent us from having to refactor naming in the future if token providers change again. Referencing generalized properties might also be more descriptive as to why ksm cares and tests those cases, instead of having them hide behind terms like "uuid", "pki", or "ferent".

[0] https://github.com/openstack/keystonemiddleware/blob/master/keystonemiddleware/tests/unit/client_fixtures.py#L119-L135