feat: add MiniMax as first-class LLM provider

[octo-patch]

Summary

Add MiniMax cloud models as a built-in LLM provider, following the established ProviderTrait pattern used by OpenAI, Anthropic, DeepSeek, and other providers.

Changes

• New MiniMaxProvider (refact-agent/engine/src/providers/minimax.rs) — full ProviderTrait implementation with:
  • OpenAI-compatible chat completions API at api.minimax.io
  • API key management with environment variable support ($MINIMAX_API_KEY) and masked display
  • Provider schema with smart links to API key page
  • Case-insensitive model filter regex ((?i)^minimax-)
  • Custom model support with configurable pricing/capabilities
• Pricing — M2.7 ($1.00/$4.00), M2.5 ($0.80/$3.00), M2.5-highspeed ($0.80/$3.00) per 1M tokens
• Registry — Registered in PROVIDER_NAMES and create_provider() factory
• README — Updated feature lists to mention MiniMax

Models

Model	Context	Use Case
MiniMax-M2.7	1M tokens	Latest flagship model
MiniMax-M2.5	204K tokens	Previous generation
MiniMax-M2.5-highspeed	204K tokens	Optimized for speed

Test Plan

• 20 unit tests covering all provider functionality
• Provider name and display name
• Wire format (OpenAI-compatible)
• Model filter regex matching
• Settings apply/export with API key masking
• Runtime building (enabled/disabled states)
• Pricing lookups for all models
• Custom model management
• Schema validation
• Clone boxing
• All tests pass: cargo test --bin refact-lsp providers::minimax