sensors: stop _measure, update_state as part of close

[ddkohler]

_measure often follows the following syntax:

def _measure(self):
    while True:
        try:
            ...
        except Exception as e:
            continue
    return measure_dict

If a daemon uses the close method while looping (e.g. usb resources are told to close), _measure may try to initiate communication after comms close, which triggers an infinite loop of exceptions. As a result, the daemon never closes, and client restart commands fail. It's a nasty form of failure because I'm not sure many daemon devs would realize this could happen. It is not 100% reproducible either (I am seeing it ~20% of the time on my particular daemon).

I propose we give is-sensor a close method that stops the looping. But I'm not sure this is possible, and I'm also not sure it's a great solution. Daemon devs would have to make sure that daemons use the trait close method prior to any hardware disconnect commands.

[ddkohler]

Summary of discussion with @ksunden

• Implementing this close method on the trait side would not just fix the issue; daemon devs would need to implement super() within their close methods. So this close implementation might see minimal usage.
• The sample _measure code above, which can have problems on restart, can be avoided by ensuring that all paths through the while loop (including exception paths) encounter an async await command, which will allow the daemon runner to close and avoid the infinite loop.

In light of this, we have decided that this issue will be addressed within individual daemons and will not be a trait implementation.

Personally, I found this issue a useful lesson on how to implement async commands carefully (e.g. even consider failures).