[Feature Request] Add 3 missing tools: image_to_image / play_video / upload_file

Background

The official MiniMax API already supports the following capabilities, but minimax-mcp-js v0.0.17 does not expose them as MCP tools:

• Image-to-Image (/v1/image/editing or similar endpoint) — transform an existing image with a text prompt + reference image(s)
• Video playback — local playback helper for downloaded .mp4 files (currently play_audio says "Does not support video")
• File upload — local file upload helper, useful as input for voice_clone, image_to_video, etc.

Why it matters

These are listed in multiple third-party / community docs (CSDN service docs, ModelScope MCP page) as part of the official MiniMax MCP feature set, so users expect them. Without them:

1. Users on Claude Code / Cursor / Windsurf cannot do iterative image editing (must use external tools)
2. Users cannot preview generated videos locally — they have to switch to Finder/Quick Look after every generate_video / image_to_video call
3. Users have to manually upload reference files to a CDN before calling image_to_video or voice_clone

Proposed tools

1. image_to_image

• Inputs: image_1 (reference), optional image_2..5, optional subject_image_*
• Arguments: prompt (required), aspect_ratio, imageCount
• Maps to: MiniMax /v1/image/editing or similar

2. play_video

• Inputs: path (local .mp4 path)
• Uses macOS open / Linux xdg-open / Windows start to launch the system default player (same pattern as play_audio but for video)

3. upload_file

• Inputs: path (local file path), optional purpose (string, e.g. voice_clone, image_ref)
• Returns: hosted URL usable as input to other tools

Request

Please consider adding these three tools in the next minor release (v0.0.18+). Happy to submit a PR if maintainers agree on the API surface — just point me at the right branch / contribution guide.

Repo cross-references that mention these tools but link to the older Python MiniMax-MCP repo:

• https://blog.csdn.net/rengang66/article/details/159322917
• https://modelscope.cn/mcp/servers/@MiniMax-AI/MiniMax-MCP

Start by locating the existing play_audio tool and the MCP tool registration or entry points in minimax-mcp-js. Verify the MiniMax image-editing and file-upload APIs before defining their inputs and outputs, and compare the platform-specific playback approach with the existing audio helper. Done means all three tools are exposed with documented behavior and coverage for their supported inputs and platforms.