docs(docker): add section on connecting to local inference servers (vLLM, Ollama)

[Magicray1217]

Summary

Adds a comprehensive guide for connecting Dockerized Hermes to local inference servers like vLLM and Ollama.

What's added

A new section Connecting to local inference servers (vLLM, Ollama, etc.) in the Docker docs page, covering:

• Docker Compose networking (recommended) — both services on the same network, using container names as hostnames
• Standalone Docker run — \host.docker.internal\ for macOS/Windows, --network host\ for Linux
• Connectivity verification — how to test from inside the container
• Ollama example — specific config for Ollama

Why

Users frequently hit networking issues when connecting Docker Hermes to local inference servers (#12308). The main gotchas:

1. Using \localhost\ instead of the container name or \host.docker.internal\
2. Mismatched --served-model-name\ vs \model:\ config value
3. Trailing slash in \�ase_url\
4. Server bound to \127.0.0.1\ instead of \

[mxnstrexgl]

✓ Automated scan: docs only. api_key: "none" is example config, not real credentials.

[alt-glitch]

Related to #7965 (open PR for same Docker-to-host LLM guide) and #7963 (the docs issue). May be competing PRs.