Reasoning-model UX: better empty-response error, higher default budget, live link

- When a model streams reasoning but no story text, return a specific,
  actionable error (raise Max output tokens / set a reasoning cap / use a
  non-reasoning model) instead of the opaque "empty response".
- Bump default max_output_tokens 400 -> 800: 400 truncated scenes and
  left reasoning models with no room after thinking. Affects new settings
  rows; existing users keep their value.
- README: add a "Try it live" link to the Render deployment.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TEpWMjnfPqzZ13nPMoGhs5

README.md

@@ -5,6 +5,10 @@ with your own AI model. Create scenarios, play open-ended adventures where an LL
 world, and extend the engine with **JavaScript scripts compatible with real AI Dungeon
 scripting**.
 
+> ### ▶️ Try it live: **[ai-dnd-1gmp.onrender.com](https://ai-dnd-1gmp.onrender.com)**
+> Play a demo scenario as a guest — no sign-up, no API key needed. (Hosted on Render's free
+> tier, so the first load after it's been idle takes ~30–60s to wake up.)
+
 Built with FastAPI + SQLite on the backend and React (Vite) on the frontend. Works with **any
 OpenAI-compatible endpoint**: Ollama and LM Studio locally, or OpenRouter / OpenAI / Groq / vLLM
 in the cloud — endpoint, key, and model are all runtime settings, and OpenRouter's free-tier