This is inspired by #1140 (comment). Quite a long time ago, I noticed that we weren't really handling scanner stop events when a scanner stops before we actually request it to. (This is somewhat similar to how we have a disconnect event for devices except in that case, the callback is called even if we requested the disconnect or not.) There haven't been any reported problems due to not having such a feature until now, so it has never been addressed until now.

So far, this PR just implements the platform-specific parts to call a common scanner backend method when an unrequested scanner stop event occurs. The idea is to use this as part of the context manager so that it will cancel the task if an early stop happens.

async with BleakScanner():
    await asyncio.sleep(10)  # BleakError("scanner stopped early") will be raised here
    # e.g. if something unexpected happened, like Bluetooth is turned off

We probably also need some sort of API for anyone using start and stop without a context manager (although maybe we don't want to support this and require a context manager if you want the callback to do something?).

Code hasn't been tested yet on any platform, so expect bugs.

cc: @bojanpotocnik