v2: Refactor and Test DeviceCollector Error Handling

[callumforrester]

Changes:

• Add tests to verify that DeviceCollector handles exceptions generated by connect()
• Add tests to verify that exceptions from devices failing to connect are logged
• Change error logging so that timeouts and exceptions from devices are logged in a uniform way
• Add comments to error handling code
• Make pre-existing tests that check the logs consistent with new tests

[callumforrester]

Some additional issues have come out of this work
#1152
#1153

[rosesyrett]

It looks like the main thing you're changing here is how the error gets logged when a device fails to connect, either because of an exception or because of a timeout error.

Seems sensible

[coretl]

My aim would be the following:

• If Device.connect() raises, then report its traceback in full
• If Device.connect() doesn't complete within timeout, cancel it, if it raises NotConnected then report the message text (which says which child signals are not connected) without traceback, otherwise report its traceback in full

This PR seems to report traceback for NotConnected errors too, do we want that?

[callumforrester]

@coretl To get to where you want, I think we would first have to fix #1153
This PR doesn't change the current behaviour except to make the log formatting more consistent. The information logged is the same.
I suggest that sorting out exactly what is logged under which circumstances should be a separate PR. Possibly also a separate issue pulling together your comment and #1153

[coretl]

Superceded in ophyd-async