fix(linux): stop the takeover firing on non-events with two devices connected

[RamfiAogusto]

Found while using the DID hook with an iPhone and a Linux laptop connected to the same AirPods Pro. Four defects, all in media_controller.rs, that make the takeover fire when the user did nothing — and with a second device connected, firing means pulling the audio away from a device they were listening on.

1. The audio card lookup does not wait

Taking the connection back makes the earbuds renegotiate, and the card drops out of the sound server for a moment. get_audio_device_index looked exactly once, so a takeover landing in that window found nothing and gave up.

The failure mode is worse than a missing profile: ownership has already been taken from the other device by then, so the audio ends up nowhere. Playback stops where it was working and never starts here.

taking ownership and activating a2dp
hijacking connection by asking AirPods
ERROR No matching Bluetooth card found for MAC CC:22:FE:54:1B:B1

Now it polls for up to three seconds.

2. Ear detection treats "unknown" as "removed"

old_all_out is .all() over the previous reading, which is vacuously true for an empty list. LibrePods reinitialises on every reconnect, so the first report of an already-worn pair reads as buds just inserted:

AirPods connected: ... initializing
Audio Source: <phone>, type: Media
Ear Detection Status: [InEar, InEar]
Ear Detection - old_in_ear_data: [], new_in_ear_data: [true, true]
... 11s later ...
Media playback started, taking ownership

The buds were never removed. The first reading is now recorded as a baseline.

3. Playback state has the same shape of bug

is_playing starts false as a placeholder, so anything already playing when the listener starts reads as playback that just began. A reconnect while media is open takes the audio. The first poll is now a baseline too.

4. Losing ownership feeds back into taking it

Losing ownership pauses local players and drops the audio profile; players tend to resume once the profile returns. That resume was indistinguishable from the user pressing play, so it took ownership back from the device that had just claimed it — which pauses this side again:

03:09:47  activated A2DP, hijacking
03:09:48  Audio Source: <laptop>, Media
03:09:49  Lost ownership, pausing media and disconnecting audio
03:09:50  Audio Source: <phone>, Media
03:09:50  taking ownership and activating a2dp
03:09:51  Audio Source: <laptop>, Media
03:09:59  Audio Source: 00:00:00:00:00:00, None

The two devices trade the audio until neither is playing. i_paused_the_media already existed for exactly this but was never set on that path.

Testing

Ubuntu 26.04, GNOME 50.1 Wayland, AirPods Pro 3, iPhone and laptop both connected via the DID hook (DeviceID = bluetooth:004c:... in BlueZ main.conf).

Before: restarting the service while the phone was playing took the audio within seconds, and the two devices traded it until both went silent.

After: a restart with media open leaves the phone playing untouched, and pressing play locally still takes the audio as intended.

One limitation is unchanged and out of scope here: if the other device is still actively playing, the earbuds may answer the hijack with OwnsConnection: 00 and keep it there. That is the earbuds arbitrating, not this code.