Refresh MiniMax LiteLLM template with global endpoint, M2.7, and video input

[octo-patch]

Reason: Refresh the LiteLLM MiniMax template to declare the global endpoint, MiniMax-M2.7, and MiniMax-M3 video input support.

Changes

• Update the commented MiniMax block in litellm_config.yaml:
  • Switch the default api_base from the CN-only endpoint (https://api.minimaxi.com/v1) to the global endpoint (https://api.minimax.io/v1), and document both the global and CN OpenAI-compatible endpoints in the comment so users can pick the right region.
  • Add a MiniMax-M2.7 model entry (text-only) alongside MiniMax-M3.
  • Declare supports_video_input: true on MiniMax-M3 so its video input capability is reflected.

Verification

• Confirmed the diff keeps the existing commented-config style and indentation consistent with the surrounding Kimi, DeepSeek, and Z.AI blocks.
• No executable tests exist for this commented template file; the change is a documentation/template refresh only.