Currently, when pycyphal encounters an error in one of its background asyncio tasks (e.g. a task receiving messages in the background), it simply calls logger.exception. This is fine for debugging with a human looking at the logs, but insufficient for production environments where you need to react to unexpected problems — there is no way of knowing an exception happened except by reading the logs.

This PR introduces set_internal_error_handler, which can be used to register a global hook into pycyphal. This hook will receive all background exceptions when they happen and can react to them as needed (e.g. by raising a fatal error). When no hook is registered, the behavior is unchanged — exceptions are simply logged.