fix(py): type evaluator callbacks as BaseEvalDataPoint

[a2105z]

Summary

• Change Python EvaluatorFn to take BaseEvalDataPoint (test_case_id: str) instead of BaseDataPoint (test_case_id optional), matching the JS evaluator API.
• Construct a BaseEvalDataPoint in the stepper (UUID fill-in) before calling the callback, without mutating the dataset BaseDataPoint.
• Update first-party evaluators (genkit-evaluators, google-genai Vertex evaluators) and the evaluators sample to use the required id directly.
• Add regression tests for id fill-in / non-mutation and callback typing.

Fixes #5991

cc @jeffdh5

Test plan

• pytest evaluator helper + veneer eval + genkit-evaluators tests (10 passed)
• Confirm a custom evaluator can use datapoint.test_case_id without or '' under the typechecker

[gemini-code-assist]

Code Review

This pull request refactors the evaluator typing system to ensure that evaluator callbacks receive a BaseEvalDataPoint with a guaranteed test_case_id instead of a BaseDataPoint where the ID is optional. A helper function _as_eval_datapoint has been introduced to automatically generate a UUID for missing test case IDs without mutating the original dataset. Corresponding updates have been made across various plugins, tests, and samples. There are no review comments, so I have no feedback to provide.

[a2105z]

@jeffdh5 ready for review when you have a moment.

This mirrors the JS EvaluatorFn / BaseEvalDataPoint split, and I reconfirmed the evaluator tests pass locally (helper + veneer eval + genkit-evaluators). Happy to adjust anything.