Python Schemas: error encoding from generated user types with an "id" field

This reveals a couple of issues:

• We probably shouldn't store the schema id in an "id" attribute on the type
• We should just treat the named tuple instance as a tuple when reading rather than calling getattr for each schema field