fix: Remove evaluation metric key from schema which failed on some LLMs

[jsonbailey]

Note

Medium Risk
Changes the structured output schema and parsing contract for judge evaluations, which can affect compatibility with existing providers/models and downstream consumers of the judge response shape. Test updates reduce risk but runtime behavior may differ if any integrations still return the old evaluations object.

Overview
Updates judge structured-output handling to use a fixed schema { "evaluation": { "score", "reasoning" } } rather than generating a dynamic evaluations object keyed by evaluation_metric_key.

Judge now always builds the schema (no key parameter), parses only the single evaluation payload, and marks evaluations as failed when that object is missing/invalid; tests are updated accordingly and dynamic schema helper methods are removed.

^{Written by Cursor Bugbot for commit 916df2a. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 3 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}