feat(search): frontend content search (text + warnings) + manga warning banner

Surfaces the analysis pipeline in the UI:

- api clients: searchPages() + PageSearchItem/ContentWarning types;
  listMangas cwInclude/cwExclude; MangaDetail.content_warnings.
- /search gains a content-search mode: free-text over OCR+scene and
  tri-state content-warning toggles (include/exclude), driven by URL
  params (text, cw_include, cw_exclude) and the new /me/page-search
  endpoint. Tag browsing is preserved when no content filter is active.
- manga detail renders a deduped content-warning banner.

Tests: vitest param serialization (searchPages CSV/text/cw, listMangas
cw); Playwright text-search + cw-toggle flows (route-mocked). svelte-check
+ build clean; full vitest (258) + search e2e (7) green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

frontend/src/routes/search/+page.ts

@@ -4,6 +4,10 @@ import {
     listMyPageTags,
     listTaggedChapters,
     listTaggedMangas,
+    searchPages,
+    CONTENT_WARNINGS,
+    type ContentWarning,
+    type PageSearchItem,
     type PageTagSummary,
     type TaggedChapterAggregate,
     type TaggedMangaAggregate,
@@ -16,6 +20,17 @@ export const ssr = false;
 type View = 'pages' | 'chapters' | 'mangas';
 type Order = 'desc' | 'asc';
 
+/** Parse a CSV of content warnings, keeping only valid vocabulary values. */
+function parseWarnings(raw: string | null): ContentWarning[] {
+    if (!raw) return [];
+    return raw
+        .split(',')
+        .map((s) => s.trim().toLowerCase())
+        .filter((s): s is ContentWarning =>
+            (CONTENT_WARNINGS as string[]).includes(s)
+        );
+}
+
 /**
  * Loader for the /search page. URL is the source of truth — refresh
  * or share a link lands the user on the same view.
@@ -37,21 +52,46 @@ export const load: PageLoad = async ({ url }) => {
         viewParam === 'chapters' || viewParam === 'mangas' ? viewParam : 'pages';
     const order: Order = url.searchParams.get('order') === 'asc' ? 'asc' : 'desc';
 
+    // Content-search inputs (OCR/scene text + content-warning filters).
+    const text = (url.searchParams.get('text') ?? '').trim();
+    const cwInclude = parseWarnings(url.searchParams.get('cw_include'));
+    const cwExclude = parseWarnings(url.searchParams.get('cw_exclude'));
+    // A positive filter (text or an included warning) is required for the
+    // backend; cw_exclude alone can't drive a search.
+    const contentSearch = text !== '' || cwInclude.length > 0;
+
     const empty = {
         authenticated: true,
         tag,
         view,
         order,
+        text,
+        cwInclude,
+        cwExclude,
+        contentSearch,
         distinct: [] as PageTagSummary[],
         pages: [] as TaggedPageItem[],
         chapters: [] as TaggedChapterAggregate[],
         mangas: [] as TaggedMangaAggregate[],
+        results: [] as PageSearchItem[],
         total: 0,
         error: null as string | null
     };
 
     try {
         const distinct = await listMyDistinctPageTags(undefined, 100);
+
+        // Content search takes precedence over tag browsing.
+        if (contentSearch) {
+            const r = await searchPages({
+                text: text || undefined,
+                cwInclude,
+                cwExclude,
+                limit: 100
+            });
+            return { ...empty, distinct, results: r.items, total: r.page.total ?? 0 };
+        }
+
         // No tag selected → just the chip cloud.
         if (!tag) return { ...empty, distinct };