docker: add linux/arm64 to CUDA cuda12 build platforms

[Copilot]

Make sure to read the contributing guidelines before submitting a PR

Extends the CUDA 12 Docker image build to target both linux/amd64 and linux/arm64, enabling multi-arch images for ARM64 hosts with NVIDIA GPUs (e.g., AWS Graviton + GPU, Jetson).

• .github/workflows/docker.yml: cuda cuda12 matrix entry platforms changed from "linux/amd64" to "linux/amd64,linux/arm64"

Original prompt

The user wants to build the linux/arm64 docker image for CUDA in their forked repository codeanker/llama.cpp.
Currently, the .github/workflows/docker.yml workflow only builds linux/amd64 for the CUDA configuration.

I need to modify .github/workflows/docker.yml to add linux/arm64 to the platforms list for the CUDA matrix entry.

Original line (around line 43):

- { tag: "cuda cuda12", dockerfile: ".devops/cuda.Dockerfile", platforms: "linux/amd64", full: true, light: true, server: true, free_disk_space: true,  runs_on: "ubuntu-22.04", cuda_version: "12.4.0", ubuntu_version: "22.04" }

New line:

- { tag: "cuda cuda12", dockerfile: ".devops/cuda.Dockerfile", platforms: "linux/amd64,linux/arm64", full: true, light: true, server: true, free_disk_space: true,  runs_on: "ubuntu-22.04", cuda_version: "12.4.0", ubuntu_version: "22.04" }

This change enables the cross-platform build for ARM64. The workflow already handles dynamic repository owner naming for the push target, so no changes are needed there.

This pull request was created from Copilot chat.

---

🔒 GitHub Advanced Security automatically protects Copilot coding agent pull requests. You can protect all pull requests by enabling Advanced Security for your repositories. Learn more about Advanced Security.