Per-operation context + initial payload

[AndreyMZ]

This pull request adds two features:

1. Per-operation context - allows sharing any data between different calls to Subscription.publish for the same subscription (and the same connection).
2. Initial payload - allows sending an arbitrary "initial" response to a client as soon as it is subscribed (before any call to Subscription.broadcast).

I use them like so (in schema.py):

import channels_graphql_ws
import graphene
from channels_graphql_ws.operation_context import OperationContext
from graphene import ResolveInfo

from .task import Task

class TaskLogChanged(channels_graphql_ws.Subscription):
    class Arguments:
        id = graphene.Argument(graphene.NonNull(graphene.ID))

    content = graphene.Field(graphene.NonNull(graphene.List(graphene.NonNull(graphene.String))))
    begin = graphene.Field(graphene.NonNull(graphene.Int))
    end = graphene.Field(graphene.NonNull(graphene.Int))

    class Meta:
        initial_payload = None

    @staticmethod
    def subscribe(root: None, info: ResolveInfo, id: str) -> Optional[List[str]]:
        task = Task(id)
        return [task.uuid]

    @classmethod
    def trigger(cls, task: Task) -> None:
        cls.broadcast_sync(group=task.uuid, payload=None)

    @staticmethod
    def publish(payload: None, info: ResolveInfo, id: str) -> Union[TaskLogChanged, object]:
        assert isinstance(info.context, OperationContext)
        saved_cursor: int = info.context.operation_context.get('cursor', 0)

        task = Task(id)
        content, new_cursor = task.get_log(saved_cursor)

        if new_cursor == saved_cursor:
            assert len(list(content)) == 0
            return TaskLogChanged.SKIP
        else:
            info.context.operation_context['cursor'] = new_cursor
            return TaskLogChanged(content=content, begin=saved_cursor, end=new_cursor)

[fdev]

@stmarkevich @sergey-komissarov @prokher is there something preventing this pull request from being merged? I'm eagerly waiting for the initial payload feature to become available. 🙏