Refine build matrix and CUDA architecture specifications

[LoserCheems]

Update the architecture matrix and default CUDA architectures to include only relevant versions for wheel builds. Refine Python and Torch version specifications for improved compatibility and remove unnecessary ARM64 OS from the build matrix.

[Copilot]

Pull Request Overview

This PR refines the build configuration by removing unnecessary architectures and platform combinations to streamline the wheel build process. The changes focus on supporting only the most relevant CUDA architectures and removing ARM64 builds.

• Reduced CUDA architecture support from 6 architectures (80, 86, 89, 90, 100, 120) to 2 essential ones (80, 90)
• Removed ARM64 OS support from the build matrix
• Updated Python version support to exclude 3.13

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
setup.py	Updates default CUDA architectures from 6 to 2 key versions
.github/workflows/publish.yml	Removes ARM64 OS and Python 3.13 from build matrix, reduces architecture specifications

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}