refactor: R12 S3 — invert scheduler->routes, publish cross-module privates

Layering cleanup, no behaviour change (ruff clean backend+scripts, pytest 193):

- Invert the scheduler->routes dependency: reset_demo_state (+ its demo-seed
  helpers) move from routes/admin.py to a new service module app/demo.py.
  Both routes/admin.py and scheduler.py now import it from app.demo, so the
  scheduler no longer reaches into the routes layer — its lazy import (the
  routes<->scheduler cycle workaround) is gone, hoisted to a top-level import.

- Publish the private symbols that were imported across module boundaries (the
  underscore convention was lying): feed._serialize -> serialize,
  feed._to_tsquery_str -> to_tsquery_str, videos._insert_stubs -> insert_stubs,
  videos._stub_from_playlist_item -> stub_from_playlist_item,
  worker._uploader_url -> uploader_url, plex._shared._item_or_404 ->
  item_or_404. Scoped renames — the 5 other route modules keep their own truly
  private _serialize helpers.

- _release_asset moves from routes/downloads.py to app/downloads/service.py as
  the public release_asset (asset bookkeeping is a service concern); the routes
  call it via service.release_asset.

backend/app/sync/videos.py

@@ -44,7 +44,7 @@ def parse_dt(value: str | None) -> datetime | None:
         return None
 
 
-def _insert_stubs(db: Session, rows: list[dict]) -> int:
+def insert_stubs(db: Session, rows: list[dict]) -> int:
     """Idempotently insert video stubs; returns the number of newly inserted rows.
 
     Uses RETURNING so the count is exact (ON CONFLICT DO NOTHING leaves rowcount
@@ -76,7 +76,7 @@ def apply_rss_feed(db: Session, channel: Channel, entries: list[dict]) -> int:
     """Insert new video stubs from an already-fetched feed and stamp the channel's last RSS
     time. DB-only — must run on the session's own thread; the network fetch
     (fetch_channel_feed) is kept separate so callers like run_rss_poll can parallelise it."""
-    inserted = _insert_stubs(db, entries)
+    inserted = insert_stubs(db, entries)
     channel.last_rss_at = _now()
     db.add(channel)
     db.commit()
@@ -84,7 +84,7 @@ def apply_rss_feed(db: Session, channel: Channel, entries: list[dict]) -> int:
 
 
 # --- Backfill (uploads playlist; costs quota) ---
-def _stub_from_playlist_item(item: dict, channel_id: str) -> dict | None:
+def stub_from_playlist_item(item: dict, channel_id: str) -> dict | None:
     content = item.get("contentDetails", {})
     snippet = item.get("snippet", {})
     video_id = content.get("videoId")
@@ -120,7 +120,7 @@ def backfill_channel_recent(db: Session, yt: YouTubeClient, channel: Channel) ->
         data = yt.get_playlist_items(channel.uploads_playlist_id, page_token)
         stopped_mid_page = False
         for item in data.get("items", []):
-            stub = _stub_from_playlist_item(item, channel.id)
+            stub = stub_from_playlist_item(item, channel.id)
             if stub is None:
                 continue
             if stub["published_at"] and stub["published_at"] < cutoff:
@@ -147,7 +147,7 @@ def backfill_channel_recent(db: Session, yt: YouTubeClient, channel: Channel) ->
     channel.recent_synced_at = _now()
     channel.backfill_cursor = next_cursor
     channel.backfill_done = done
-    inserted = _insert_stubs(db, collected)
+    inserted = insert_stubs(db, collected)
     db.add(channel)
     db.commit()
     return inserted
@@ -166,11 +166,11 @@ def backfill_channel_deep(
         rows = [
             s
             for s in (
-                _stub_from_playlist_item(it, channel.id) for it in data.get("items", [])
+                stub_from_playlist_item(it, channel.id) for it in data.get("items", [])
             )
             if s is not None
         ]
-        total += _insert_stubs(db, rows)
+        total += insert_stubs(db, rows)
         page_token = data.get("nextPageToken")
         channel.backfill_cursor = page_token
         if not page_token: