feat(issue-2): Project scaffold, rubric schema, and base evaluator class

[Nidhi18-git]

Summary

Adds the foundational structure for Issue #2 — AI-based evaluation of 21st-century skills.

What's Included

• issue-2-ai-skill-evaluator/ directory with clean project structure
• data/sample_rubric.json — a 4-level rubric schema for skill assessment (creativity, critical thinking, problem solving, agency)
• src/evaluator.py — abstract BaseEvaluator class with EvaluationResult dataclass; to be extended for each model
• src/data_utils.py — data loading helpers (skeleton)
• requirements.txt — full dependency list for VLM fine-tuning stack
• Detailed README.md with model candidates, milestones, and project structure

Motivation

The repo currently has no code. This PR creates the shared foundation that future contributions (data pipeline, model benchmarking, fine-tuning) can build on.

Related Issue

Closes #2 (partially — this is the scaffold, not the full implementation)

Next Steps (what I'll tackle next)

• Model benchmarking notebook with zero-shot evaluation on sample rubric
• Data preprocessing pipeline for student artifact images
• Cost analysis script comparing open-source models vs Gemini

[TarunaJ2006]

I explored the project scaffold and reviewed the proposed workflow for rubric-based evaluation of 21st-century skills. The repository structure and modular design for evaluation, data utilities, and rubric handling are well organized and provide a strong foundation for future development. I am currently understanding the repository architecture, project goals, and planned implementation stages in more detail. I would like to contribute through documentation improvements, setup guidance, and preprocessing-related utilities that can help improve usability, reproducibility, and contributor onboarding for the evaluation pipeline.