[codex] Add remote Ollama service support

[Yan-ShiBo]

Summary

Add support for choosing Ollama models from either the local machine or a LAN server reached through an SSH tunnel.

Changes

• Add remote Ollama source parsing and routing in the server, including source-aware model selection, health output, keepalive, and unload behavior.
• Add a tray Remote Service dialog for server name, IP, SSH port, username, password, Ollama host, and Ollama port.
• Run the remote service dialog on a background thread so its text fields stay clickable.
• Make tray quit force-exit after cleanup so the background app does not get stuck when exiting.
• Populate the web model selector with local and remote model options from the server health response.
• Add Paramiko as the SSH dependency and include it in setup verification.
• Document LAN remote Ollama setup in README and Chinese instructions.
• Add focused Python and userscript tests for remote source routing and UI model option merging.

Validation

• python -m pytest -q -> 161 passed, 15 subtests passed
• node --test tests/userscript-core.test.cjs -> 23 passed

Notes

tray_settings.json is intentionally not included because it can contain local server credentials.