[Bug]: Suggest upgrading base image from Debian 12 to Debian 13 for GLIBC compatibility on ARM64

## QwenPaw Version

latest (Docker official image)

## Description

When running local LLM via QwenPaw on ARM64, the model fails to start. The backend returns error: "llama.cpp 服务器在准备就绪前退出".

Root cause: Official Docker image uses Debian 12 (GLIBC 2.36), while the aarch64 version of latest llama.cpp requires GLIBC 2.38 or higher, causing the service to exit abnormally.

**Related PR(s):**

**Security considerations:** None

## Component(s) Affected

- [x] Core / Backend (app, agents, config, providers, utils, local_models)
- [ ] Console (frontend web UI)
- [ ] Channels (DingTalk, Feishu, QQ, Discord, iMessage, etc.)
- [ ] Skills
- [ ] CLI
- [ ] Documentation (website)
- [ ] Tests
- [ ] CI/CD
- [x] Scripts / Deploy

## Environment

- **QwenPaw version:** latest (Docker)
- **OS:** Debian 12 (Docker base image)
- **Install method:** Docker official image `agentscope/qwenpaw:latest`
- **Python version (if applicable):** 3.11.2
- **Architecture:** ARM64 (aarch64)

## Steps to Reproduce

1. Deploy QwenPaw with official Docker image on ARM64 server
2. Install llama.cpp through QwenPaw built-in local model installation
3. Load and start a local LLM model

## Actual vs Expected

- **Actual:** Failed to start model, error: {"detail":"llama.cpp 服务器在准备就绪前退出"}
- **Expected:** Local LLM model can be loaded and run normally on ARM64

## Logs / Screenshots
{"detail":"llama.cpp 服务器在准备就绪前退出"}

## Additional Notes

Please upgrade the Docker base image from Debian 12 to Debian 13. Debian 13 provides GLIBC 2.38 by default, which can resolve the dependency issue and allow llama.cpp to run normally on ARM64.