feat: add MiniMax as first-class LLM provider

[octo-patch]

Summary

Add MiniMax AI as a first-class LLM provider with full integration across the ChatAny stack.

Supported Models

• MiniMax-M2.7 - Latest flagship model with 204K context
• MiniMax-M2.5 - Balanced model for general use
• MiniMax-M2.5-highspeed - Optimized for low-latency responses

Changes (16 files, 869 additions)

• Constants (constant.ts): ServiceProvider.MiniMax, ModelProvider.MiniMax, ApiPath.MiniMax, base URL, model list
• Client Platform (platforms/minimax.ts): MiniMaxApi with streaming SSE via OpenAI-compatible API
• Server Proxy (api/minimax/[...path]/route.ts): Edge Runtime proxy with auth and model filtering
• Auth (auth.ts): MiniMax API key injection
• Server Config (server.ts): MINIMAX_API_KEY and MINIMAX_URL env vars
• Access Store (access.ts): MiniMax URL/key fields + isValidMiniMax() validation
• Settings UI (settings.tsx): MiniMax config in provider dropdown
• Locales (en.ts, cn.ts): English and Chinese locale strings
• Factory Dispatch (api.ts): ClientApi and getClientApi dispatch
• Docs: README.md and README_EN.md updates
• Tests: 29 unit + integration tests with vitest

Configuration

Set MINIMAX_API_KEY env var or configure via Settings > Custom Config > MiniMax.

Test Plan

• 16 unit tests: constants, models, enums, API config
• 13 integration tests: client API path, message extraction, factory dispatch
• All 29 tests pass
• Manual: Select MiniMax provider in settings and verify chat works