feat: add MiniMax as alternative LLM provider

[octo-patch]

Summary

• Add MiniMax as a direct LLM provider alongside OpenRouter
• Users can switch between providers in Settings > LLM Provider
• MiniMax models: M2.7, M2.7-highspeed, M2.5, M2.5-highspeed (1M / 204K context)
• Full tool-calling support with execute_command — all 32 Flipper actions work
• Temperature clamping (0,1] and <think> tag stripping for MiniMax models
• Settings UI: provider dropdown, conditional API key field, per-provider model list

Changes

File	What
LlmProvider.kt	Provider enum (OPEN_ROUTER, MINIMAX)
MiniMaxClient.kt	OpenAI-compatible client with rate limiting, retry, tool calling
OpenRouterClient.kt	Expose shared constants for MiniMaxClient reuse
VesperAgent.kt	Route chat() to selected provider; keep OpenRouter for vision
SettingsStore.kt	MiniMax API key + provider preference persistence
SettingsScreen.kt	Provider selector, conditional API key field, dynamic model list
SettingsViewModel.kt	Provider/MiniMax state management
README.md	MiniMax models table, setup instructions, architecture diagram

Test plan

• 28 unit tests covering thinking-tag stripping, response parsing, model catalog, provider enum, tool-call arguments, edge cases
• 3 integration tests (require MINIMAX_API_KEY env var, auto-skipped otherwise)
• Manual: switch provider to MiniMax, enter API key, verify chat works
• Manual: verify OpenRouter still works after switching back
• Manual: verify model dropdown updates per provider

10 files changed, 1,458 additions, 62 deletions