Add MiniMax as LLM provider for target and controller models

[octo-patch]

Summary

• Add MiniMax as a first-class LLM provider alongside OpenAI, Anthropic, Google, XAI, and Ollama
• MiniMax uses an OpenAI-compatible API at https://api.minimax.io/v1 with models like MiniMax-M2.7 (1M context) and MiniMax-M2.5
• Can be used as both target and controller model via --target-model-type minimax / --controller-model-type minimax

Changes

• promptmap2.py: Add minimax to CLI choices, API key validation, client initialization (OpenAI SDK with custom base_url), chat completion handling with think tag stripping for reasoning models
• README.md: Add MiniMax to provider list, API key docs, controller model recommendations
• tests/test_minimax_provider.py: 19 unit tests covering API key validation, client init, prompt sending, think tag stripping, error handling, CLI arg parsing
• tests/test_minimax_integration.py: 3 integration tests (chat completion, system prompt defense, controller evaluation)

Usage

export MINIMAX_API_KEY=your-minimax-key

# Test MiniMax model as target
python3 promptmap2.py --target-model MiniMax-M2.7 --target-model-type minimax

# Use MiniMax as controller for evaluating another model
python3 promptmap2.py --target-model llama2:7b --target-model-type ollama   --controller-model MiniMax-M2.7 --controller-model-type minimax

Test plan

• 19 unit tests pass (mock-based, no API key needed)
• 3 integration tests pass (real MiniMax API calls)
• Existing provider behavior unchanged