fix: #36183 add auto-generate workflow/chatflow functionality

[duongynhi000005-oss]

Fixes #36183

Summary

Adds the ability to automatically generate chatflows or workflows from a natural language description using an LLM.

Changes

• Backend API endpoint: POST /workflow-generate that accepts a description and app mode, then uses an LLM to generate a complete workflow graph (nodes + edges)
• LLM prompt: A detailed system prompt that instructs the LLM on available node types, graph structure, and layout rules
• Frontend service: generateWorkflow() function to call the new endpoint
• Tests: Unit tests covering success, markdown fence stripping, error handling, and invalid JSON cases

How it works

1. User provides a natural language description of the desired workflow (e.g., "A chatbot that retrieves knowledge from a database and answers questions")
2. The system sends this to the configured LLM with a structured prompt describing all available node types and their schemas
3. The LLM returns a valid workflow graph JSON
4. The frontend can use this graph to populate the workflow canvas

What was tested

• Python syntax validation of all modified files
• Unit tests for the generate_workflow method (success, markdown fences, invoke errors, invalid JSON, missing fields)