Add Copilot Chat MCP support

[sahilk873]

This update makes the existing ShortKit-ML MCP server usable from GitHub Copilot Chat in VS Code without changing the detector code itself.

The gap in issue #3 was not the shortcut-detection API. The repo already exposed the full MCP tool surface, but it did not ship a VS Code workspace config or documented Copilot Chat flow. As a result, Copilot users had no checked-in path to discover and launch the shortkit-ml server from the repo.

This change adds a tracked .vscode/mcp.json using the same shortcut_detect.mcp_server entry point as the Claude Code/Desktop setup, documents the VS Code Agent-mode workflow in the final MCP documentation section, and updates the README MCP section to mention GitHub Copilot in VS Code. I also added a small regression test that keeps the VS Code MCP config aligned with the root project config.

Validation:

• pytest -q tests/test_mcp_server.py tests/test_copilot_mcp_config.py
• python3 -m json.tool .vscode/mcp.json
• python3 -m json.tool .mcp.json

Notes:

• This is a repo-side Copilot Chat integration via MCP, not a GitHub App extension. That matches the current documented Copilot/MCP path and keeps the server on a maintained integration surface.
• The remaining Copilot Chat smoke test is a manual VS Code step: open the tools panel, confirm shortkit-ml appears, and run a real prompt.

Closes #3.

[sebasmos]

@sahilk873 it would be great to confirm this works with Copilot in VS Code directly not only via pytests, including the free extension, and then include it in the MCP documentation/website last section, following the same approach as the Claude Desktop example: https://criticaldata.github.io/ShortKit-ML/mcp-server/