Deploy Triton Inference Server on Railway

A step-by-step guide to deploying NVIDIA Triton Inference Server on Railway, covering both simple PyTriton deployments and advanced multi-model setups with MinIO as a model registry. Demonstrates serving a dummy AddSub model and ResNet18 on CPU, configuring model repositories, dynamic model loading/unloading via MinIO object storage, and querying models using the Triton HTTP client. Includes Dockerfiles, config files, and full client code examples.