Improve media ipc

[Ape]

This adds

media <..|play|next-player|previous-player>

and changes some logic in the mpris service so that it's a lot more convenient and behaves the way I would expect.

Type of Change

• Bug fix
• New feature

Manual Coverage

• Tested on Hyprland

Checklist

• This PR is ready for review, or it is marked as Draft.
• I read and followed the relevant guidance in CONTRIBUTING.md.
• I ran just format with clang-format v22+ installed.
• I ran the relevant build or test commands.
• I self-reviewed the changes.
• I checked for new warnings or errors.
• I will update end-user documentation after merge.
• This PR adds no new user-facing strings.
• I did not edit non-English translation files unless this PR is explicitly for translation tooling, an import/export sync, or a maintainer-requested locale change.
• I used the existing canonical names for config keys, IPC names, paths, and identifiers.

[ItsLemmy]

Review: PR #3155 — "Improve media ipc"

What it does: Adds media play, media next-player, and media previous-player IPC actions; reworks stop() to prefer Pause over Stop (and no longer dismiss the player); and reworks applyPlayerSnapshot so the active/pinned player auto-follows whichever player starts/stops playing. It also drops two pause-recovery playbackStatus = "Playing" overrides. Overall a reasonable UX improvement, but it leaves orphaned subsystems and reverts a deliberate heuristic without explanation.

Findings (most severe first)

1. mpris_service.cpp:785 (removed) — Removing the sole dismissPlayer() call orphans the entire "dismissed players" subsystem.
   stop() was the only caller of dismissPlayer(), and dismissPlayer() is the only place that inserts into m_stoppedPlayers (line 2185). After this PR m_stoppedPlayers can never become non-empty, so every isDismissed() guard in chooseActivePlayer() (lines 2062, 2071, 2091, 2100, 2108, 2115), the erases at 770/1746/2034, and dismissPlayer() itself (2179) all become unreachable dead code. Either the dismiss feature is being intentionally retired — in which case this dead machinery should be deleted in this PR — or it's an unintended loss of "stop removes the player from active rotation." Worth confirming which, and cleaning up accordingly.

2. mpris_service.cpp:559 and :1606 (removed) — Dropping playbackStatus = "Playing" reverts the no-signal-pause recovery heuristic.
   Both removed lines corrected a player that falsely reports Paused while its position keeps advancing past kPauseRecoveryMinJumpUs (a known quirk on some players/compositors). With them gone, the model leaves such a player at Paused (only the m_recentNoSignalPauseAt erase remains). Since chooseActivePlayer()'s "most recent playing" path (line 2071) only considers playbackStatus == "Playing", an actually-playing-but-stuck-Paused player drops out of active selection — the exact bug that heuristic was added to fix. This may be intentional given the new active-follow logic, but it's a deliberate heuristic being reverted with no mention in the PR; the author should confirm intent.

3. mpris_service.cpp (new } else if (clearedPinnedPlayerPreference) tail of applyPlayerSnapshot) — Unreachable branch.
   clearedPinnedPlayerPreference is only ever set inside the becamePlaying or transitionedFromPlaying paths, both of which require playbackStatus to differ between existing->second and info — which guarantees existing->second != info, so the outer if always wins. The else if (the existing->second == info case) can never run with clearedPinnedPlayerPreference == true. Harmless defensive code, but dead; can be dropped.

Design note (not a bug)

The new becamePlaying clear logic resets an explicit user pin (media next-player) as soon as any other player starts playing while the pinned one isn't playing. That matches the "active follows playback" intent, but it does mean an explicit pin to a paused player is silently overridden by background playback — worth a sentence in the PR/commit so it's a documented choice rather than a surprise.