[Skila1]

Summary

Failed and stuck tracks currently call do_next() in a way that can skip unrelated queued songs. This change introduces a single queue consumer: Lavalink TrackEnd reasons go through one PlaybackSession, and QUEUE_ADVANCE moves exactly one logical item.

• Failed or stuck items advance once (that item only). There is no retry in this PR.
• Public APIs (Player.current, Queue.get / tracks / history) still return Track, not an internal queue wrapper.
• Notices, health IPC, websocket CLOSE handling, and optional stuck retry are not included.

Tests

Vocard-function tests on this branch (not Lavalink/YouTube plugin behavior):

• TrackEnd(loadFailed) / stuck → exactly one advancement
• generic replaced does not consume the next song
• skip / stale play completion does not double-advance
• public queue helpers still return Track

Also ran upstream python tests/runner.py (language packs + language usage + these playback tests) and python -m compileall voicelink cogs.

Notes

This replaces the oversized main-targeted PR with a smaller beta slice. My personal fix/playback-reliability fork is unchanged (retries/notices/health stay on there).