Back-off retries for requests to `/3pid/onbind` against homeservers after a successful bind are not tested

[anoadragon453]

#552 fixed a bug where failed attempts to call PUT /_matrix/federation/v1/3pid/onbind by Sydent were not rescheduled properly (an exception was raised, and rescheduling the failed call thus never happened).

We should test that calls to /3pid/onbind are correctly retried in a back-off fashion if they fail.

Relevant code:

sydent/sydent/threepid/bind.py

Lines 203 to 219 in c09fe7a

	def _notifyErrback(
	self, assoc: Dict[str, Any], attempt: int, error: Union[Exception, str]
	) -> None:
	"""
	Handles errors when trying to send an association down to a homeserver by
	logging the error and scheduling a new attempt.
	:param assoc: The association to send down to the homeserver.
	:param attempt: The number of previous attempts to send this association.
	:param error: The error that was raised when trying to send the association.
	"""
	logger.warning(
	"Error notifying on bind for %s: %s - rescheduling", assoc["mxid"], error
	)
	self.sydent.reactor.callLater(
	math.pow(2, attempt), self._notify, assoc, attempt + 1
	)

[matrixbot]

For your information, this issue has been copied over to the Element fork of sydent: element-hq/sydent#554