fix: replace speechSynthesis announce with real audio clip via /api/tts-announce

speechSynthesis is silently muted on iOS Safari and Chrome Android after
the audio session ends (onended), so chapter announcements never played.

Fix:
- Add GET /api/tts-announce backend endpoint: streams a short TTS clip
  for arbitrary text without MinIO caching (backend/internal/backend/)
- Add GET /api/announce SvelteKit proxy route (no paywall)
- Add announceNavigatePending/announcePendingSlug/announcePendingChapter
  to AudioStore
- Rewrite onended announce branch: sets audioStore.audioUrl to the
  announcement clip URL so the persistent <audio> element plays it;
  the next onended detects announceNavigatePending and navigates
- 10s safety timeout in case the clip fails to load/end

ui/src/routes/api/announce/+server.ts

@@ -0,0 +1,39 @@
+import { error } from '@sveltejs/kit';
+import type { RequestHandler } from './$types';
+import { backendFetch } from '$lib/server/scraper';
+
+/**
+ * GET /api/announce?text=...&voice=...&format=...
+ *
+ * Thin proxy to backend GET /api/tts-announce.
+ * No paywall — this is a short announcement clip (a few words), not chapter audio.
+ * No MinIO caching — the backend streams the clip directly.
+ */
+export const GET: RequestHandler = async ({ url }) => {
+	const text = url.searchParams.get('text') ?? '';
+	if (!text) error(400, 'text is required');
+
+	const qs = new URLSearchParams();
+	qs.set('text', text);
+
+	const voice = url.searchParams.get('voice');
+	if (voice) qs.set('voice', voice);
+
+	const format = url.searchParams.get('format') ?? 'mp3';
+	qs.set('format', format);
+
+	const backendRes = await backendFetch(`/api/tts-announce?${qs}`);
+
+	if (!backendRes.ok) {
+		error(backendRes.status as Parameters<typeof error>[0], 'TTS announce failed');
+	}
+
+	return new Response(backendRes.body, {
+		status: 200,
+		headers: {
+			'Content-Type': backendRes.headers.get('Content-Type') ?? 'audio/mpeg',
+			'Cache-Control': 'no-store',
+			'X-Accel-Buffering': 'no'
+		}
+	});
+};