Add support for custom scalars

[Ambro17]

[ethe]

In my opinion, there are several objectives we need to implement:

• [ ] Allow user to declare scalar types
• [ ] custom scalar types can be printed as a schema: https://graphql.org/learn/schema/#scalar-types
• [ ] Allow client transfer string values and load them from string to Python types (Datetime, UUID, etc.) which declared in scalar types

[ethe]

Maybe we should create a new metaclass called ScalarType and a root class called Scalar, then users could create their custom scalar types as objects of this metaclass by inheriting Scalar. There are some use cases:

from pygraphy import ScalarType

class Scalar(metaclass=ScalarType):
    # defined in pygraphy, implement some default method
    ...

class UUID(Scalar):
    def load(self, string: str) -> pyUUID:
        return pyUUID(string)

    def dump(self) -> str:
        return str(self.value)  # value is defined in `Scalar` class maybe

then UUID can be used as a normal type when declaring models:

class Foo(pygraphy.Object):
    id: UUID

[Ambro17]

Should I solve the first three items on this issue?

[ethe]

Yeah, I am glad to hear that.