Hello.

As discussed in #892, we could filter out the already transmitted downlinks from the queue flush on the call to enqueue. That allows previously sent downlinks from being acked once the device communicates back.

I kept the existing flush_for_dev_eui function as-is and created another flush_non_pending_for_dev_eui function to be called by enqueue as the former is also used in other methods like flush_queue, activate where it sounds more reasonable to flush all downlinks.

Thank you.