Improve server-client communication error handling

[MetRonnie]

In #6499 we added a new field to the trigger GraphQL mutation, and this unfortunately broke the ability for cylc trigger at 8.4 to work on workflows running in 8.3 (I shall term this "inter-version server-client comms").

This PR does not fix that (we don't think a fix is feasible at this stage).

An additional problem is that the error handling for server-client comms is broken - that is what this PR fixes. And it attempts to ensure we have defined format for server-client comms which should reduce the chance of breaking inter-version server-client comms in future.

Check List

• I have read CONTRIBUTING.md and added my name as a Code Contributor.
• Contains logically grouped changes (else tidy your branch by rebase).
• Does not contain off-topic changes (use other PRs for other changes).
• No dependency changes
• Tests are included (or explain why tests are not needed).
• Changelog entry included if this is a change that can affect users
• No docs needed
• If this is a bug fix, PR should be raised against the relevant ?.?.x branch.

[MetRonnie]

For future/historical reference, I originally tried mocking the poll method of client.poller, i.e.

mp.setattr(client.poller, 'poll', Mock())

but this resulted in pytest hanging near the end of running all integration tests; some strange interaction due to the ZMQ poller using threading or something, dunno.

Anyway, this monkeypatching is needed to get client.poller.poll() to return a truthy value so the socket can receive the mock response

cylc-flow/cylc/flow/network/client.py

Lines 298 to 300 in dd468d6

	# receive response
	if self.poller.poll(timeout):
	res = await self.socket.recv()