fix: freeze set during ticks iter in async hub

[claudinoac]

Relates to #1774

With redis broker, when the redis server reaches timeout and celery worker is running a long cpu-intensive task, the transport tick automatically disconnects, and that's expected.

The connection error is thrown by the redis-py lib, but before that happens, the redis transport layer is auto-disconnected.
Instead of propagating the ConnectionError, the async hub gets its ticks set changed in the middle of the for loop that executes its callbacks, resulting in an error such as RuntimeError: Set changed size during iteration.
This happens because one of these callbacks belongs to the transport layer itself.

This solution basically freezes those ticks sets so they keep their consistency until the ConnectionError is thrown, which allows the celery worker to automatically recover itself, rather than just teardown due to an unrecoverable error like that one above.

[Nusnus]

Thank you and sorry for the delay!

[claudinoac]

@Nusnus Thank you!