fix: keep snake case in openapi schema token authentication

Part of frontend#75 (closed)

While working on frontend#75 (closed), I found out that the OpenAPI schema camelcases the token authentication endpoint. That's because the transformation to camelcase is done on the schema level and the function is not aware of the view overriding the parser_classes and renderer_classes. Because the client SDKs are generated from the OpenAPI schema, currently the clients are not working with the token authentication endpoint, because we use snake_case there. This MR fixes that, by wrapping the camelize_serializer_fields function and keeping the TokenRequest and TokenResponse schemas in snake_case.