A new KD method, FSDP/FSDP2 support, and experiment tracking

New method

• Add SKD method (PRs #538, #541)

New features

• Extend data parallel support for FSDP/FSDP2 (PR #537)
• Support trackio and wandb (PR #543)
• Add accumulation mode with stack option (PR #536)

Improvements

• Use a more robust eval for DDP (PR #532)
• Remove teacher I/O caching (PR #539)
• Extend example log_analyzer.py (PR #544)
• Add yaml examples (PR #542)
• Refactor (PR #526)

Tests and CI

• Add loss test cases with fixtures (PR #535)
• Replace setup.py with pyproject.toml (PR #534)
• Add a Copilot instruction file and update GitHub Actions (PRs #528, #530, #531, #533)

Documentation

• Update README's (PR #540)
• Update acknowledgments (PRs #527, #529)
• Update year (PR #523)

Misc

• Update version and notebooks (PRs #522, #545)
• Update dependencies (PR #539)