ci: support CUDA 13.0 Docker builds and publishing

[lvhan028]

No description provided.

[Copilot]

Pull request overview

This PR updates CI and internal build guidance to support building and publishing LMDeploy Docker images for CUDA 13.0 alongside CUDA 12.8, using canonical Docker build targets (cu130 / cu128) while keeping public tags (cu13.0 / cu12.8) consistent.

Changes:

• Update Docker CI workflows to build/test and publish CUDA 13.0 (cu130) and CUDA 12.8 (cu128) variants.
• Adjust Docker publish tagging logic to use separate “build target” vs “tag suffix” values.
• Refresh manywheel builder/wheel README and the internal docker-build skill documentation with CUDA 13.0 examples.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
builder/manywheel/README.md	Updates documentation for building/pushing builder images and building CUDA 13.0 wheels.
.github/workflows/test_docker.yml	Tests Docker builds across cu130 and cu128 build targets.
.github/workflows/docker.yml	Publishes Docker images for CUDA 13.0/12.8 using canonical build targets and versioned tags.
.claude/skills/docker-build/SKILL.md	Updates internal build instructions to map public CUDA variants to canonical Docker stage targets.

Suppressed comments (1)

builder/manywheel/README.md:22

• This section describes selecting a CUDA version, but the example uses GPU_ARCH_VERSION (the variable consumed by build_lmdeploy_builder.sh). Adding a short note here would reduce confusion for users who look for CUDA_VERSION.

For custom builds with specific manylinux and CUDA versions:

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.