Update RAPIDS Custom Docker Documentation from CUDA 12.9.1 to CUDA 13.1.1

[jayavenkatesh19]

Updated all documentation to use CUDA 13.1.1 to align with RAPIDS 25.12's CUDA 13 support.

• Updated example Dockerfiles to use CUDA 13.1.1 base images
• Changed pip package suffix from cudf-cu12 to cudf-cu13
• Updated all CUDA package versions in the custom base image guide to match official NVIDIA CUDA 13.1.1 Dockerfiles
• Removed NCCL packages from pip section (not in official runtime images)
• Added ENTRYPOINT [] to pip Dockerfiles to suppress license message produced by the CUDA Runtime images.

I also tested GPU visibility, a basic cuDF test and CUDA versions by building these Docker images on the dlcluster

[ncclementi]

Should this be a good opportunity to start templating the CUDA version, to start migrating all the docs when possible to cuda 13?