Add MCP (Model Context Protocol) server support

[webhead2oo9]

Summary

This PR implements MCP (Model Context Protocol) support for the Discord bot as requested in issue #3.

What's been added

Core MCP Implementation

• MCPClient: Handles connections to MCP servers via stdio and SSE transports
• MCPManager: Singleton that manages multiple MCP server connections centrally
• MCPToolAdapter: Converts MCP tools to the bot's BaseTool interface seamlessly
• MCPResourceHandler: Handles MCP resource fetching and attachment conversion
• MCPPromptHandler: Manages MCP prompts for system instruction enhancement

Configuration Support

• Added mcpServerConfigurations section to config.json for defining MCP servers
• Extended connectors to accept mcpServers array alongside regular tools
• Added comprehensive validation for all MCP configurations

Key Features

✅ MCP tools are automatically discovered and integrated with AI models
✅ MCP resources are converted to Discord attachments
✅ Support for both stdio (local) and SSE (remote) MCP servers
✅ Works seamlessly with existing tool infrastructure
✅ Comprehensive error handling and logging
✅ TypeScript types and full build compatibility

Example Configuration

{
  "mcpServerConfigurations": {
    "filesystem": {
      "transportType": "stdio",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/directory"]
    },
    "github": {
      "transportType": "stdio",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "your-token"
      }
    }
  },
  "connectorConfigurations": {
    "ToolsOpenAIConnector": {
      "connectionOptions": {
        "tools": ["SearxingTool"],
        "mcpServers": ["filesystem", "github"]
      }
    }
  }
}

Documentation

• Added comprehensive MCP_INTEGRATION.md guide with detailed setup instructions
• Updated README.md with MCP features
• Updated template.config.json with example MCP server configurations

How it works

1. MCP servers are defined centrally in mcpServerConfigurations
2. Connectors specify which MCP servers they want to use via mcpServers array
3. On startup, the bot connects to all configured MCP servers
4. MCP tools are automatically discovered and made available to AI models
5. The tools work exactly like native tools - no special handling required

This implementation opens up exciting possibilities like filesystem access, GitHub integration, database queries, and much more through the growing MCP ecosystem.

Closes #3