feat: tiered LLM (GLM free / Claude paid) + rate limits + quota enforcement

The free tier was hemorrhaging Anthropic cost with no abuse cap (no rate
limit on /preview, Opus default in the build worker, 5-min cache TTL that
made cache-miss the common case). This switches free users to GLM, paid
users to Claude tiers, and tightens every leak found in the audit.

Backend:
- @bmm/llm: GLM provider via Zhipu's OpenAI-compatible endpoint, pickPreviewModel
  + pickBuildModel helpers, plan-aware ModelChoice
- preview-cache TTL 5min -> 24h (kills the cache-miss path)
- /v1/servers/preview: picks model from caller's plan, returns model name to UI
- /v1/servers POST: enforces SERVER_LIMITS per plan (402), rate-limits builds
- daily rate-limit on preview (5/40/150/1000) and build (3/20/100/500)
- /v1/auth/me returns plan so the wizard can show the right model name
- generator worker: GLM default, Anthropic Sonnet fallback if GLM errors

Frontend:
- Wizard fetches plan, shows "<model> is drafting the tool spec" pre-emptively,
  upgrade hint for hobby users, friendly errors for 402 / 429
- Pricing page: AI-model line per tier (Open-tier / Haiku / Sonnet / Opus),
  Team €149 -> €199, Enterprise €499 -> €999, daily-preview limit per tier
- Privacy + Security: explicit subprocessor disclosure for Anthropic (US) /
  Zhipu (CN) and which tier uses which

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

apps/web/app/(marketing)/privacy/page.tsx

@@ -36,11 +36,21 @@ const SECTIONS = [
   {
     h: 'Subprocessors',
     p: [
-      "Anthropic (generation) — only the prompt text you send. Anthropic's data-retention policy applies.",
-      'Hetzner (compute).',
-      'Backblaze (encrypted backups).',
-      'Stripe (billing).',
-      'Cloudflare (DNS + DDoS).',
+      "Anthropic, USA (Claude AI — used for prompt analysis and code generation on Pro / Team / Enterprise tiers). Only the prompt text and resulting spec are sent. Anthropic's data-retention policy applies.",
+      'Zhipu AI, China (GLM model — used for prompt analysis on the free Hobby tier only). Only the prompt text and resulting spec are sent. Upgrade to a paid tier to keep all AI processing within Anthropic (US).',
+      'Hetzner, Germany (compute).',
+      'Backblaze, EU (encrypted backups).',
+      'Stripe, Ireland (billing).',
+      'Cloudflare (DNS + DDoS protection).',
+    ],
+  },
+  {
+    h: 'AI processing per tier',
+    p: [
+      'Hobby (free): prompts are sent to Zhipu AI (GLM, China) for analysis. Choose a paid tier if your prompts contain data that must not leave the EU/US.',
+      'Pro: prompts are sent to Anthropic (Claude Haiku 4.5, USA).',
+      'Team: prompts are sent to Anthropic (Claude Sonnet 4.6, USA).',
+      'Enterprise: Anthropic (Claude Sonnet + Opus, USA) with EU-data-residency opt-in available on request.',
     ],
   },
   {