deserialization of Record objects incomplete when using rpc calls

[tarbaig]

Version Current master

Steps to reproduce

#adding.py
# based on examples/agent.py
import faust
from faust import Topic

class Add(faust.Record, serializer='json'):
    a: int
    b: int

class Result(faust.Record,serializer='json'):
    r : int
    summands : Add

class Settings:
    FAUST_KAFKA_BROCKER = 'kafka://localhost:9093'
    KAFKA_BROCKERS = ['localhost:9093']

class EcoTopic(Topic):
    def __init__(self, *args, **kwargs) -> None:
        """
        The default replication for Faust topics is broken, this class changes the default
        replication from 0 to 1 and should be used as replacement for Fausts own topics until this is fixed in the
        library
        """
        if kwargs.get("replicas") == 0:
            kwargs["replicas"] = 1
        super().__init__(*args, **kwargs)

app = faust.App(
    'add_app',
    broker=Settings.FAUST_KAFKA_BROCKER,
    Topic=EcoTopic
)

topic = app.topic('adding', value_type=Add)

@app.agent(topic)
async def adding(stream):
    async for value in stream:
        yield Result(value.a + value.b,value)

#send_to_agent.py
import faust
from addition.addition_app import Add,Result, adding

class Settings:
    FAUST_KAFKA_BROCKER = 'kafka://localhost:9093'
    KAFKA_BROCKERS = ['localhost:9093']


async def sum_ask(a, b) -> None:
    add_object = Add(a=a, b=b)
    result = (await adding.ask(add_object))
    print(f"result={result}")
    return result

if __name__ == '__main__':
    r = compute_sum_ask(0,0)
    assert isinstance(r, Result)

Expected behavior

Return value should be of class Result

Actual behavior

Return value is plain json/dict

##Fix
When fulfilling reply promises the return value should be wrapped in a call to maybe_model

##Details
Faust uses some internal store to Register all created Subclasses of Record, when deserializing some special fields in the
deserialized data are used to detect that this should be actually converted to a Record. The corresponding logic is in maye_model.