feat: add MiniMax Cloud API as alternative prompt enhancement provider

[octo-patch]

Summary

• Add MiniMax Cloud API as an alternative LLM provider for prompt enhancement (PE) in HunyuanImage-3.0, alongside the existing DeepSeek/Tencent Cloud integration
• Introduce --llm-provider CLI flag (deepseek/minimax) and --sys-prompt-type flag (universal/text_rendering) for flexible prompt rewriting configuration
• MiniMax uses an OpenAI-compatible API at https://api.minimax.io/v1 with models like MiniMax-M2.7 (1M context window) — users only need pip install openai and a MINIMAX_API_KEY

Changes

File	Description
PE/minimax_client.py	New MiniMaxClient class with OpenAI-compat API, temperature clamping (0,1], think-tag stripping, retry logic
run_image_gen.py	Multi-provider dispatch via --llm-provider flag, new --sys-prompt-type arg (fixes missing arg reference)
README.md	MiniMax installation step, usage example, CLI arguments table
tests/test_minimax_provider.py	31 unit tests + 3 integration tests

Motivation

The current PE module only supports DeepSeek via Tencent Cloud's LKEAP SDK, which requires Tencent Cloud credentials. Adding MiniMax as an alternative:

• Provides a simpler setup path (standard OpenAI SDK + API key)
• Gives users choice of LLM backend for prompt enhancement
• MiniMax-M2.7 offers a 1M token context window suitable for complex prompt engineering

Test plan

• 31 unit tests covering: think-tag stripping, client init, recaption, arg parsing, provider dispatch, system prompt compatibility
• 3 integration tests with real MiniMax API (universal prompt, text rendering, M2.5 model)
• All 34 tests passing
• Manual test: python3 run_image_gen.py --prompt 'test' --rewrite 1 --llm-provider minimax (requires GPU for full image generation)

Usage

# With MiniMax PE
export MINIMAX_API_KEY="your_key"
python3 run_image_gen.py \
    --model-id ./HunyuanImage-3 \
    --prompt "A brown and white dog running on grass" \
    --rewrite 1 \
    --llm-provider minimax