feat(phase5): optional generative RAG — ask your lectures, with citations

Strictly opt-in, gated, with the deterministic features as the always-present floor:
- GenerationEngine: WebLLM (Qwen2.5-1.5B, WebGPU, CDN-loaded) + BYO-key cloud
  (OpenAI-compatible); native stub. Pure grounding prompt builder (4 tests).
- rag.askLectures: retrieve Phase-1 hits -> grounded prompt -> answer with
  citations; refuses when nothing relevant; falls back to search-only when no
  engine is available. Never sends raw audio/transcripts — only question + snippets.
- aiStore (BYO key persisted in localStorage on web), Ask screen (answer +
  tappable citation chips that jump to the audio + honest "verify" disclaimer),
  Settings AI section (engine status + bring-your-own-key form).

279 tests green, 0 tsc errors, web export builds. ROADMAP phases 0-5 complete.

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

src/lib/generation/engineImpl.native.ts

@@ -0,0 +1,32 @@
+// NATIVE on-device generation stub. Running a local LLM on device (e.g. via
+// llama.rn / MLC's native runtime) is a Phase 5 follow-up; until then the native
+// webllm engine reports unavailable so callers transparently fall back to the
+// search-only path (or a BYO-key cloud model). It NEVER produces a fake answer.
+//
+// This module is selected by Metro for native (.native.ts) and is never imported
+// by any vitest test.
+
+import type { GenerationEngine } from './engine';
+
+const NOT_AVAILABLE = 'On-device generation is not available on native yet';
+
+export const webllm: GenerationEngine = {
+  kind: 'webllm',
+  label: 'On-device (Qwen2.5-1.5B)',
+
+  async isAvailable(): Promise<boolean> {
+    return false;
+  },
+
+  async loadModel(): Promise<void> {
+    throw new Error(NOT_AVAILABLE);
+  },
+
+  isLoaded(): boolean {
+    return false;
+  },
+
+  async generate(): Promise<string> {
+    throw new Error(NOT_AVAILABLE);
+  },
+};