Validate sample and rubric JSON with Pydantic

[JeremyZeng77]

Summary

• Adds Pydantic validation models for sample and rubric JSON payloads.
• Keeps load_sample() and load_rubric() returning dictionaries so existing CLI/GUI/eval callers keep working.
• Reports invalid payloads with source path and field path details for easier fixture debugging.
• Adds tests that validate all existing sample/rubric fixtures and assert helpful errors for invalid payloads.

Fixes #4.

Validation

• python -m pytest tests/test_data_schema_validation.py tests/test_eval.py tests/test_language_coverage.py -q (11 passed)
• python -m ruff check src tests (passed)
• python -m compileall src tests (passed)
• python -m pytest -q (71 passed)
• git diff --check (passed)

Bounty

Claimed in #4 and MergeOS Claim Token #1 with required repos starred.

[JeremyZeng77]

Light follow-up: this PR is still ready for review. It keeps the existing loader dict interface intact and includes validation evidence in the PR body (
uff, compileall, and full pytest -q passing locally). Happy to adjust if maintainers prefer a different schema strictness level.