Add GPU-based HPA for Triton Inference Server

[uzunenes]

Added a tutorial/guide for deploying NVIDIA Triton Inference Server on Kubernetes with Horizontal Pod Autoscaling based on GPU utilization metrics using DCGM Exporter and Prometheus.

The guide covers:

• Triton Inference Server deployment
• GPU metrics collection with DCGM Exporter
• Prometheus integration for custom metrics
• HPA configuration for automatic scaling based on GPU utilization