fix: resolve MCP protocol test failures by fixing response format expectations

- Fixed test-helpers.ts to correctly wrap executeTool responses in MCP format
- Updated all tests to expect correct response structures:
  - list_nodes returns {nodes: [...], totalCount}
  - search_nodes returns {query, results: [...], totalCount, mode?}
  - list_ai_tools returns {tools: [...]}
  - list_tasks returns {totalTasks, categories: {...}} or {category, tasks: [...]}
- Fixed property expectations (nodeType instead of name, etc.)
- Reduced failing tests from 67 to 7

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

tests/integration/mcp-protocol/test-helpers.ts

@@ -51,15 +51,9 @@ export class TestableN8NMCPServer {
     // Call tool handler
     this.server.setRequestHandler(CallToolRequestSchema, async (request) => {
       try {
+        // The mcpServer.executeTool returns raw data, we need to wrap it in the MCP response format
         const result = await this.mcpServer.executeTool(request.params.name, request.params.arguments || {});
         
-        // Convert result to content array if needed
-        if (Array.isArray(result) && result.length > 0 && result[0].content) {
-          return {
-            content: result[0].content
-          };
-        }
-        
         return {
           content: [
             {