feat: add MiniMax-M2.7 models and fix legacy MiniMax API config

[octo-patch]

Summary

• Upgrade MiniMax models from legacy abab6.5s-chat / abab7-chat-preview to the latest MiniMax-M2.7 and MiniMax-M2.7-highspeed (204K context window)
• Fix API base URL from deprecated api.minimax.chat to api.minimax.io for all MiniMax entries
• Fix temperature from 0 to 0.01 — MiniMax API requires temperature in (0.0, 1.0]
• Extend API key detection in OpenAIWrapper to match MiniMax-M model name pattern alongside legacy abab, with fallback to MINIMAX_API_KEY env var
• Applied consistently to both VLMEvalKit/ and VLMEvalKit_ov/

Usage

# Set your MiniMax API key
export MiniMax_API_KEY=your-key-here
# or
export MINIMAX_API_KEY=your-key-here

# Run evaluation with MiniMax-M2.7
python run.py --model MiniMax-M2.7 --data MMBench_DEV_EN

# Run with highspeed variant
python run.py --model MiniMax-M2.7-highspeed --data MMBench_DEV_EN

Changes

File	Change
VLMEvalKit/vlmeval/config.py	Add M2.7 + M2.7-highspeed entries, update abab URL/temp
VLMEvalKit/vlmeval/api/gpt.py	Extend key detection for MiniMax-M pattern
VLMEvalKit_ov/vlmeval/config.py	Same config updates (mirror)
VLMEvalKit_ov/vlmeval/api/gpt.py	Same key detection update (mirror)
VLMEvalKit/tests/test_minimax_provider.py	17 unit + 3 integration tests

Test plan

• 17 unit tests pass: config registry, API base URL, model names, temperature constraints, key detection
• 3 integration tests pass: text generation and working() check against live MiniMax API
• Legacy abab models remain registered and functional (backward compatible)