feat: add MiniMax as LLM provider preset (default M3)

[octo-patch]

Summary

Add MiniMax as a built-in LLM provider preset with MiniMax-M3 set as the default model, giving users one-click access to a 512K context window, 128K max output, and image input support via an OpenAI-compatible API.

Changes

• Add MiniMax preset to llm_presets.py (name, base_url, model = MiniMax-M3, api_key_url)
• Add temperature clamping in LLMService for MiniMax API (requires temperature in (0.0, 1.0])
• Strip <think> tags from structured output responses for MiniMax compatibility
• Update README.md and README_EN.md to mention MiniMax in supported models and recommended solutions
• Add MiniMax preset comment in config.example.yaml
• Add 21 unit tests and 3 integration tests

Why M3 as default

MiniMax-M3 is the latest model: 512K context window, up to 128K output, and image input support. Users can still switch to MiniMax-M2.7 or MiniMax-M2.7-highspeed by overriding the model field.

Test Plan

• 21 unit tests pass (preset validation, temperature clamping, mock LLM calls, utility functions)
• 3 integration tests pass against live MiniMax API (basic call, temperature clamping, structured output)
• Manual test: select MiniMax preset in Web UI, verify auto-fill of base_url and model

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

PR Bot seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[Jefsky]

Hi @octo-patch! PR #114 (#114) fixes a critical issue in this PR: the base_url uses api.minimax.io which cannot fetch the model list. The correct URL is api.minimax.chat/v1. You may want to update this PR to use the correct base URL so it can be merged. Thanks!