feat(admin): LLM tip quality dashboard — per-model/prompt/kind breakdowns

/admin/reward-analytics now surfaces served count, reaction rate, and avg
reward grouped by llm_model, prompt_version, and tip_kind — closing the
loop so model/prompt iterations in M2 are legible next to the bandit
policy view. Data comes from the tip_scores columns added in ffdf707 and
tip_feedback.reward_milli; bandit-only tips show as "(bandit-only)".

Closes #92.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

services/api/src/test/db.ts

@@ -2,11 +2,13 @@
  * Creates an isolated in-memory SQLite DB with the full schema applied.
  * Use this in tests instead of the shared `db` singleton.
  */
-import Database from 'better-sqlite3';
+import Database, { type Database as BetterSqlite3Database } from 'better-sqlite3';
 import { drizzle } from 'drizzle-orm/better-sqlite3';
 import * as schema from '../db/schema.js';
 
-export function makeTestDb() {
+type DrizzleDb = ReturnType<typeof drizzle<typeof schema>>;
+
+export function makeTestDb(): DrizzleDb & { rawSqlite: BetterSqlite3Database } {
   const sqlite = new Database(':memory:');
   sqlite.pragma('foreign_keys = ON');
 
@@ -138,7 +140,10 @@ export function makeTestDb() {
     );
   `);
 
-  return drizzle(sqlite, { schema });
+  const db = drizzle(sqlite, { schema });
+  // `sqlite` is exposed as `rawSqlite` so tests that mock `../db/index.js`
+  // can provide the same `{ db, rawSqlite }` shape as the prod module.
+  return Object.assign(db, { rawSqlite: sqlite });
 }
 
 export type TestDb = ReturnType<typeof makeTestDb>;