fix(eval): retry judge timeouts with the request options - #8121

Please ensure you have read the contribution guide before creating a pull request.

Link to Issue or Description of Change

1. Link to an existing issue (if applicable):

• Closes: Eval LLM judges do not retry client-side TimeoutError / google.auth.exceptions.TimeoutError despite HttpRetryOptions #8075

Problem:
Eval judges attach HttpRetryOptions, but a client TimeoutError still fails the sample on the first try. TimeoutError is not in google-genai's aiohttp transient set, and a non-Gemini judge never reads the retry options. LlmAsJudge, HallucinationsV1Evaluator, and PerTurnUserSimulatorQualityV1 do not retry the timeout themselves, so one timed-out sample marks the invocation NOT_EVALUATED.

Solution:
read_judge_response_with_retry builds a tenacity AsyncRetrying from the request's HttpRetryOptions (attempts, backoff, jitter) and logs each retry. The three judges above call it. Other errors are not retried. When attempts are used up the sample stays not evaluated. User simulators are not in this change, and google-genai's transient list is not modified.

Testing Plan

Unit Tests:

• I have added or updated unit tests for my change.
• All unit tests pass locally.

test_read_judge_response_retries_a_timeout covers both timeout types. test_read_judge_response_stops_after_the_attempt_count stops at attempts=2. test_read_judge_response_does_not_retry_other_errors keeps a RuntimeError to one call. test_evaluate_invocations_retries_a_judge_timeout checks LlmAsJudge uses the helper.

pytest tests/unittests/evaluation/test_retry_options_utils.py tests/unittests/evaluation/test_llm_as_judge.py tests/unittests/evaluation/test_hallucinations_v1.py tests/unittests/evaluation/simulation/test_per_turn_user_simulation_quality_v1.py

106 passed. mypy on the four source files: no issues.

Manual End-to-End (E2E) Tests:

I did not run adk eval against a live model. The tests use a mocked judge.

Checklist

• I have read the CONTRIBUTING.md document.
• I have performed a self-review of my own code.
• I have commented my code, particularly in hard-to-understand areas.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.
• I have manually tested my changes end-to-end.
• Any dependent changes have been merged and published in downstream modules.