feat(feed): broaden search to a weighted document (title + keywords + queries + description)

Makes the local search behave more like YouTube's — finding videos by the
uploader's own keywords or the query that surfaced them, not only words in the
title. A DB-generated, weighted search_vector (migration 0032) replaces the
title-only FTS index:
- keywords: the creator's snippet.tags (free — already in the snippet we fetch),
  stored on enrich.
- search_terms: distinct live-search queries that surfaced the video (across all
  users), appended by the search route — folds YouTube's relevance into local
  search (a video YT returned for a query becomes findable by it even without a
  title match), the user's own idea.
- description (truncated) for broad recall on the existing catalog.
Weighted title(A) > keywords+queries(B) > description(C) so ts_rank keeps title
hits on top. A plain GIN index on the generated column guarantees index use (no
expression/param matching). Verified on localdev: recall 146->213 for one query;
7 'eurovision' hits via the document but not the title; index scan confirmed.

backend/app/models.py

@@ -3,6 +3,7 @@ from datetime import date, datetime
 from sqlalchemy import (
     BigInteger,
     Boolean,
+    Computed,
     Date,
     DateTime,
     Float,
@@ -14,6 +15,7 @@ from sqlalchemy import (
     UniqueConstraint,
     func,
 )
+from sqlalchemy.dialects.postgresql import TSVECTOR
 from sqlalchemy.orm import Mapped, mapped_column, relationship
 
 from app.db import Base
@@ -237,6 +239,14 @@ class Video(Base, TimestampMixin):
     topic_categories: Mapped[list | None] = mapped_column(JSON)
     default_language: Mapped[str | None] = mapped_column(String(16))
     detected_language: Mapped[str | None] = mapped_column(String(16))
+    # Creator-supplied keyword tags (snippet.tags joined by spaces) — fetched for free with the
+    # snippet we already request. Indexed into search_vector so the feed search matches the
+    # uploader's own keywords, not just words in the title.
+    keywords: Mapped[str | None] = mapped_column(Text)
+    # Accumulated distinct live-search queries that surfaced this video (across all users) —
+    # folds YouTube's own relevance judgement into the local search: a video YouTube returned
+    # for "ti amo magyarul" becomes findable by that query locally even if its title lacks it.
+    search_terms: Mapped[str | None] = mapped_column(Text)
     is_short: Mapped[bool] = mapped_column(
         Boolean, default=False, server_default="false", index=True
     )
@@ -272,6 +282,21 @@ class Video(Base, TimestampMixin):
     )
     unavailable_reason: Mapped[str | None] = mapped_column(String(24))
 
+    # Weighted full-text search document (DB-generated, never written by the app): title (A) >
+    # creator keywords + search queries (B) > truncated description (C). ts_rank honours the
+    # weights so title matches outrank description matches. Backed by a GIN index (migration
+    # 0032). The accent-insensitive `unaccent_simple` config comes from migration 0031.
+    search_vector: Mapped[object | None] = mapped_column(
+        TSVECTOR,
+        Computed(
+            "setweight(to_tsvector('public.unaccent_simple', coalesce(title, '')), 'A') || "
+            "setweight(to_tsvector('public.unaccent_simple', coalesce(keywords, '') || ' ' || "
+            "coalesce(search_terms, '')), 'B') || "
+            "setweight(to_tsvector('public.unaccent_simple', left(coalesce(description, ''), 1000)), 'C')",
+            persisted=True,
+        ),
+    )
+
     channel: Mapped["Channel"] = relationship(back_populates="videos")