Add few key points on the README + schema

[mehd-io]

Thank you for creating smallpond! While exploring the codebase for a technical deep dive article, I noticed some key architectural aspects weren't featured in the README. This PR improves that to better showcase smallpond's capabilities.

Changes

• Emphasized distributed processing
• Added Ray integration details: Explicitly mentioned Ray's role in the distributed execution model
• Clarified storage options: Specified support for local filesystem and 3FS
• Added architecture diagram: Included a mermaid flowchart visualizing high-level architecture
• Improved architecture description

Let me know your thoughts, happy to cut some parts.

[wangrunji0408]

Thanks for the PR and the detailed write-up! Here are some of my thoughts:

• The architecture diagram is clear and informative, but we prefer to keep the README concise. It would be a better fit in the project documentation under an "Architecture" section.
• Currently, S3 is not supported as a storage option. While the local filesystem is available for development and testing, production-level deployments rely on high-performance distributed file systems like 3FS.
• Smallpond actually has two backends: the built-in scheduler and Ray. We're working on making it easier for users to choose between them, so we avoid emphasizing Ray as the primary dependency.

[mehd-io]

@wangrunji0408

Thanks for the PR and the detailed write-up! Here are some of my thoughts:

• The architecture diagram is clear and informative, but we prefer to keep the README concise. It would be a better fit in the project documentation under an "Architecture" section.

Good point, moved to the docs.

• Currently, S3 is not supported as a storage option. While the local filesystem is available for development and testing, production-level deployments rely on high-performance distributed file systems like 3FS.

Sorry, I overlooked that on my side. It’s fixed now.

• Smallpond actually has two backends: the built-in scheduler and Ray. We're working on making it easier for users to choose between them, so we avoid emphasizing Ray as the primary dependency.

I understand, but I feel it’s okay to highlight Ray for now, as it is the primary target (and the easiest?) for people to use at a production level. When it's easier to use a built-in scheduler, we can add more docs on these two paths.

I also noticed that we didn’t reference the hosted docs later in the README, so I updated the URLs. I initially missed the “Docs” badge, and I assume most people will as well. To make it more visible, I highlighted it again at the top of the README. I also recommend adding it under the “About” section as a Website link.