feat: add eval-schemas, eval-fixtures, eval-notebook checks

[adnanrhussain]

Stacked on #120 (ahussain/eval-config-checks). Review/merge that first; this PR's diff is against that base.

What

Three new checks, completing the evaluator-artifact validation set started in #120:

Check	Validates
eval-schemas	each input_schema.json / output_schema.json is a well-formed JSON Schema document (meta-validation)
eval-fixtures	fixtures.json against the shared evals/_schemas/fixtures.schema.json, and binds each case's input to the evaluator's input_schema and each expected field to its output_schema property
eval-notebook	where an evaluator ships an example notebook, confirms its code loads config.json + the declared prompt files from disk rather than hardcoding them

All register into the same harness; CI runs them via python scripts/check.py.

Notes

• eval-fixtures degrades gracefully if an evaluator's schema is itself malformed (that's eval-schemas' job to report) — guarded with check_schema so it never crashes the run.
• eval-notebook is a lightweight heuristic (string-level), not notebook execution.
• All 8 current evaluators pass all five checks. Negative-tested each new check to confirm it actually fires.

[Copilot]

Pull request overview

Adds three new evaluator artifact validation checks to the existing scripts/check.py harness, rounding out schema/fixture/notebook consistency checks introduced in the prior stacked PR.

Changes:

• Adds eval-schemas to meta-validate input_schema.json / output_schema.json as JSON Schema documents.
• Adds eval-fixtures to validate fixtures.json structure and bind fixture input/expected to the evaluator’s schemas.
• Adds eval-notebook to heuristically verify example notebooks load config.json and prompt files from disk, plus updates documentation/registry.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
scripts/README.md	Documents the new checks in the harness README.
scripts/checks/eval_schemas.py	New check for JSON Schema meta-validation of evaluator schemas.
scripts/checks/eval_fixtures.py	New check for fixtures schema validation + per-evaluator binding.
scripts/checks/eval_notebook.py	New check ensuring notebooks reference config/prompts on disk (heuristic).
scripts/checks/base.py	Simplifies shared_schema_path() helper to a single-argument form.
scripts/checks/init.py	Registers the new checks in the harness.
README.md	Links to scripts/ harness and recommends running scripts/check.py --fix.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[czi-fsisenda]

Nice! Great way to enforce correct use of the json patterns for the notebooks.
If we're including fixtures jsons, should we find a way to converge the json pattern with the contract tests?