fix: preserve expected_outcome in conversational golden conversion

[aerosta]

Summary

ConversationalGolden.expected_outcome was being dropped when goldens were converted to ConversationalTestCase. This could cause conversational metrics that require expected_outcome to fail validation or skip evaluation after conversion.

This change preserves expected_outcome in convert_convo_goldens_to_convo_test_cases(), so converted conversational test cases keep the same evaluation inputs as the source golden.

Changes

• preserve expected_outcome when converting ConversationalGolden to ConversationalTestCase
• add a regression test covering the conversion path

Test plan

DEEPEVAL_TELEMETRY_OPT_OUT=1 python -m pytest tests/test_core/test_datasets/test_dataset.py -q --tb=short
DEEPEVAL_TELEMETRY_OPT_OUT=1 python -m pytest tests/test_core/test_test_case/test_multi_turn/test_conversational_test_case.py -q --tb=short

[vercel]

@aerosta is attempting to deploy a commit to the Confident AI Team on Vercel.

A member of the Team first needs to authorize it.