Initial implementation of an MCP server as an extension

[Zsailer]

Summary

This PR refactors the MCP server extension to use a simplified, trait-based architecture that removes complex abstractions while providing more flexibility and better Jupyter integration.

Key Changes

• Simplified MCPServer: Direct function registration without tool provider/registry abstractions
• Trait-based Configuration: Full traitlets integration for server and extension configuration
• Dynamic Tool Loading: String-based tool specifications in format module:function
• Modern Extension API: Updated to use current Jupyter Server extension lifecycle
• Comprehensive Documentation: Updated README with usage examples and troubleshooting guide

Architecture Improvements

• Removed complex tool provider and registry classes
• MCPServer now inherits from LoggingConfigurable for proper Jupyter integration
• Extension supports configurable tool lists via mcp_tools trait
• Automatic tool registration from popular packages (jupyter-ai-tools, jupyterlab-commands-toolkit)
• Simplified async lifecycle management with proper cleanup

Configuration Example

c = get_config()

c.MCPExtensionApp.mcp_name = "Jupyter MCP Server"  
c.MCPExtensionApp.mcp_port = 8080
c.MCPExtensionApp.mcp_tools = [
    "os:getcwd",
    "jupyter_ai_tools.toolkits.notebook:read_notebook",
    "jupyterlab_commands_toolkit.tools:clear_all_outputs_in_notebook",
]

Testing

• All tests pass (35/35) with comprehensive coverage
• Updated tests to match simplified API
• Removed tests for deleted abstractions
• Added integration tests for real-world usage

Documentation

• Complete README rewrite with usage examples
• Troubleshooting guide for common issues
• API documentation for all components
• Configuration examples from minimal to advanced

This refactor maintains full backward compatibility for the MCP protocol while significantly simplifying the internal architecture and improving maintainability.

Test Plan

• All existing tests pass
• New integration tests added
• Manual testing with real Jupyter environments
• Documentation verified with examples
• Configuration validation tested