fix(docker): use Docker Hub for UBI10 CUDA base images (#1511)

## Summary

- Change UBI10 Dockerfile FROM lines from `nvcr.io/nvidia/cuda:...` to `nvidia/cuda:...`

Authors:
  - Ramakrishna Prabhu (https://github.com/ramakrishnap-nv)

Approvers:
  - Trevor McKay (https://github.com/tmckayus)

URL: #1511

Author: ramakrishnap-nv

ci/docker/Dockerfile.ubi

@@ -6,13 +6,13 @@ ARG CUDA_VER=unset
 ARG CUOPT_VER=unset
 
 # Stage 1: source for libnvrtc / libnvJitLink (not in base image)
-FROM nvcr.io/nvidia/cuda:${CUDA_VER}-runtime-ubi10 AS cuda-libs
+FROM nvidia/cuda:${CUDA_VER}-runtime-ubi10 AS cuda-libs
 
 # Stage 2: source for CUDA headers needed for CuPy/NVRTC runtime compilation
-FROM nvcr.io/nvidia/cuda:${CUDA_VER}-devel-ubi10 AS cuda-headers
+FROM nvidia/cuda:${CUDA_VER}-devel-ubi10 AS cuda-headers
 
 # Stage 3: small base — install Python + cuOpt via pip
-FROM nvcr.io/nvidia/cuda:${CUDA_VER}-base-ubi10 AS python-env
+FROM nvidia/cuda:${CUDA_VER}-base-ubi10 AS python-env
 
 ARG CUDA_VER
 ARG CUOPT_VER