µEd API support with /evaluate, /preview, and /health endpoints

[m-messer]

Summary

This PR integrates the µEd API (openapi-v0_1_0) into the base evaluation layer alongside the existing legacy API, enabling evaluation functions to serve µEd-compliant requests without breaking backward compatibility.

What changed

New µEd routing in handler.py

• POST /evaluate — validates the request against the EvaluateRequest schema, delegates to commands.evaluate_muEd, and validates the response against the EvaluateResponse schema before returning.
• POST /preview — delegates to commands.preview_muEd, returning a pre-submission feedback item.
• GET /evaluate/health — returns service capabilities (supported artefact types, formats, and feature flags).
• Legacy routes (/) are unchanged and continue to work as before.

Schema reorganisation

• Legacy JSON schemas moved under schemas/legacy/ to make room for the new µEd OpenAPI spec at schemas/muEd/openapi-v0_1_0.yml.

feedbackId made optional

• The Feedback schema previously required feedbackId. This field is now optional — evaluation functions are not required to generate IDs, and the platform is not required to consume them. This matches the intent of the µEd spec, where feedbackId is a convenience for clients that need to correlate feedback items, not a mandatory field.
• Removed hardcoded "feedbackId": "fb-1" and "feedbackId": "preSubmissionFeedback" from commands.py.
• Removed the required: [feedbackId] constraint from both the OpenAPI schema and the inline validation schema in validate.py.

Validation

• validate.py now supports OpenAPI-backed validation for µEd schemas using jsonschema.RefResolver, in addition to the existing file-based JSON Schema validators for legacy schemas.

Tests

• Full test coverage for µEd request/response validation (mued_requests_test.py, mued_responses_test.py).
• Integration-style handler tests for all three µEd endpoints (mued_handling_test.py).