Add MiniMax as a new LLM provider client (default: MiniMax-M3)

[octo-patch]

Summary

• Add MiniMaxClient extending ChatGPTClient for MiniMax's OpenAI-compatible Chat Completions API
• Default to the latest MiniMax-M3 model (512K context window, up to 128K output, image input support); keep MiniMax-M2.7 and MiniMax-M2.7-highspeed as alternatives
• Wire MiniMax into the API server, CLI, settings, and module exports as a first-class minimax client option
• Temperature clamping to MiniMax's [0, 1] range, proper user/assistant role message construction

Available models

Model	Notes
MiniMax-M3 (default)	Latest MiniMax model — 512K context, 128K max output, image input
MiniMax-M2.7	Previous generation
MiniMax-M2.7-highspeed	Previous generation, low-latency variant

Changes

File	Description
src/MiniMaxClient.js	New client extending ChatGPTClient with overridden buildPrompt for proper chat message roles; default model set to MiniMax-M3, default maxContextTokens set to 512000
bin/server.js	Add minimax case in getClient() factory
bin/cli.js	Add minimax client option with MiniMax AI label
index.js	Export MiniMaxClient
settings.example.js	Add miniMaxClient configuration section (M3 default, M2.7/M2.7-highspeed listed as alternatives)
demos/use-minimax-client.js	Usage example with M3 defaults
test/minimax-client-unit.js	Unit tests covering constructor, options, temperature clamping, buildPrompt, stop tokens (default-model assertions updated to MiniMax-M3, default maxContextTokens assertion updated to 512000, plus a new test that overriding to MiniMax-M2.7 still works)
test/minimax-client-integration.js	Integration tests targeting MiniMax-M3
README.md	Document MiniMax support with MiniMax-M3 as the default and MiniMax-M2.7 / MiniMax-M2.7-highspeed as alternatives; note the 512K context / 128K output window

Why M3 as default

MiniMax-M3 is MiniMax's latest model and supersedes the M2.x line: 512K context window, up to 128K output, and image input support. M2.7 and M2.7-highspeed are kept as user-selectable alternatives so existing deployments remain functional, while older M2.5 entries are removed since they are superseded.

Test plan

• Unit tests updated for M3 default model and 512K maxContextTokens (node --test test/minimax-client-unit.js)
• Integration tests target MiniMax-M3 (MINIMAX_API_KEY=... node --test test/minimax-client-integration.js)
• ESLint passes on all new files
• Verify clientToUse: 'minimax' works in API server mode
• Verify clientToUse: 'minimax' works in CLI mode

[octo-patch]

Pushed a fix for the CI failure: clientToUseForMessage was declared inside the try block but referenced in the error handler, so any request error would throw a ReferenceError instead of returning the proper message. Hoisted the declaration and cleaned up the lint warnings flagged by airbnb-base.