Add data organization extension and restructure Data Efficacy documentation

[dialogueeeeee]

Summary

This PR updates the DELT repository into a scalable Data Efficacy codebase and adds the ACL 2026 follow-up work, Demystifying Data Organization for Enhanced LLM Training.

Main changes:

• Adds new data ordering methods for the ACL 2026 data organization work.
• Keeps the original DELT scoring, selection, training, and evaluation pipeline intact.
• Reorganizes documentation into paper-specific pages under docs/.
• Updates the root README to introduce Data Efficacy, list supported works, and provide the shared pipeline usage.

Validation

• Checked markdown links locally.
• Checked whitespace with git diff --check.
• Full pre-training, SFT, and evaluation experiments require the DELT GPU environment.

[dialogueeeeee]

@microsoft-github-policy-service agree company="Microsoft"