feat(sources): YouTube Music search + download pipeline (§1C/§1E)

Pluggable fetch source: ytmusicapi search + yt-dlp download (cookies-file guard), DownloadJob entity/repo + DownloadService, download_task worker with exponential-backoff retries, and wired /search, /sources/{source}/search, and /downloads endpoints. Adds youtube_enabled/cookies config, yt-dlp+ytmusicapi deps, and the download_jobs.track_id migration. Snapshot also bundles in-progress storage/tracks/acoustid edits.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

app/workers/queue.py

@@ -34,6 +34,20 @@ async def enqueue(function: str, **kwargs: Any) -> str:
     return str(job.job_id)
 
 
+async def enqueue_download(job_id: uuid.UUID) -> None:
+    """Best-effort enqueue of a download job for the worker.
+
+    The job row is already persisted as ``queued``, so this is a follow-up, not a
+    barrier: if the queue is unreachable we log and move on (graceful
+    degradation) — the job stays ``queued`` and can be retried later. Deferred a
+    few seconds so the request's DB transaction commits before the worker reads
+    the row (same reason as :func:`enqueue_enrich`)."""
+    try:
+        await enqueue("download_track", job_id=str(job_id), _defer_by=3)
+    except DependencyUnavailableError:
+        log.warning("download_enqueue_failed", job_id=str(job_id))
+
+
 async def enqueue_enrich(track_id: uuid.UUID) -> None:
     """Best-effort enqueue of metadata enrichment for a freshly stored track.