Add support for typing.NewType fields

[loopj]

It is sometimes convenient to use typing.NewType when defining model fields, to provide additional static type checking.

Here's a contrived example:

Speed = NewType("Speed", float) # Speed in MPH

@dataclass
class Car:
    speed: Speed

At runtime, python effectively converts Speed straight back into a float:

>>> car = Car(100.0)
>>> car.speed
100.0

>>> type(car.speed)
<class 'float'>

But since xsdata is using get_type_hints under the hood, this returns the type hinted type directly:

>>> get_type_hints(Car)
{'speed': car.Speed}

>>> type(get_type_hints(Car)["speed"])
<class 'typing.NewType'>

>>> get_type_hints(Car)["speed"].__supertype__
<class 'float'>

This causes parsing to blow up:

xsdata.exceptions.XmlContextError: Error on Car::speed: Xml Attribute does not support typing `car.Speed`

I propose we add a built-in converter for typing.NewType types, which handles these cases by calling __supertype__.