Add multi-instance deployment and load balancing configuration

[scyyh11]

• Introduced compose.multi_instance.yaml.example for deploying multiple PaddleOCR API instances with a shared VLM inference service.
• Added nginx.multi_instance.conf for Nginx load balancing configuration to distribute requests across API instances.
• Created README.multi_instance.md to guide users on setting up and using the multi-instance deployment with Nginx.

[paddle-bot[bot]]

Thanks for your contribution!