Update skills and add conda package recipe with CI/CD support

[zhyan0603]

This pull request introduces several improvements and updates across documentation, packaging, and developer guidance for GPUMDkit. The main highlights are the addition of a Conda publishing workflow, significant enhancements to developer and agent documentation, improved help and usage messages, and updated project URLs.

Key changes include:

1. Packaging & Distribution

• Added a new GitHub Actions workflow (.github/workflows/publish-conda.yml) to automate building and publishing the Conda package for multiple platforms, including upload to Anaconda.org on tagged releases.
• Added a .gitattributes rule to enforce LF line endings for all .sh scripts, ensuring cross-platform compatibility.

2. Documentation Updates

• Updated all references and links from the old zhyan0603.github.io site to the new gpumdkit.cn domain in README.md and Scripts/README.md. [1] [2] [3]
• Updated version and date in banners and help tables to reflect the latest release (2026-07-10). [1] [2] [3] [4]
• Improved help output in README.md for clarity, including the addition of -skill and -doctor options, and clarified usage instructions. [1] [2]

3. Agent & Developer Guidance

• Overhauled AGENTS.md to clarify project architecture, agent routing rules, and reference organization. The new version emphasizes agent skill usage, documentation structure, and strict change policies.
• Updated CONTRIBUTING.md to clarify the structure of agent skills, improve code and menu input conventions, and require more robust argument parsing and help messages in Python scripts. Also, shell and Python best practices are now more explicit. [1] [2] [3] [4] [5] [6] [7]

4. Usability Improvements

• Improved menu and workflow documentation in Scripts/README.md, including more readable workflow tool menus and explicit support for new sampling modes. [1] [2]
• Added -dp2xyz (DeepMD npy → extxyz) to the main help table for better discoverability.

These changes collectively improve the maintainability, usability, and distribution of GPUMDkit, making it easier for both users and contributors to interact with the project.