MAINT Fix integration test import errors and runtime issues

[romanlutz]

Problem

Several integration test files fail to collect or run due to code issues introduced when AttackIdentifier and ScorerIdentifier were consolidated into ComponentIdentifier:

• 3 test files fail with ImportError: cannot import name 'AttackIdentifier' from 'pyrit.identifiers' — caused by tests/integration/mocks.py importing the removed AttackIdentifier
• 1 test file (test_azure_sql_memory_integration.py) fails with ImportError: cannot import name 'ScorerIdentifier' and additionally 2 tests fail at runtime because AttackResult is constructed with a plain
  dict for attack_identifier, but AttackResultEntry.init calls .to_dict() on it
• 1 test file (test_hitl_gradio_integration.py) fails with ModuleNotFoundError: No module named 'rpyc' when rpyc is not installed
• 1 source file (add_image_to_video_converter.py) causes cv2.error on headless OpenCV builds from cv2.destroyAllWindows()

Changes

tests/integration/mocks.py

• Replace AttackIdentifier import and type hint with ComponentIdentifier

tests/integration/memory/test_azure_sql_memory_integration.py

• Replace ScorerIdentifier import with ComponentIdentifier
• Update get_test_scorer_identifier() to return ComponentIdentifier with scorer-specific fields in params
• Add get_test_attack_identifier() helper returning a proper ComponentIdentifier
• Replace 6 instances of attack_identifier={"name": "test_attack"} with get_test_attack_identifier()
• Update assertions in roundtrip test to access fields via params dict

tests/integration/score/test_hitl_gradio_integration.py

• Guard rpyc-dependent imports with try/except
• Add pytestmark to skip entire module when rpyc is not installed

pyrit/prompt_converter/add_image_to_video_converter.py

• Wrap cv2.destroyAllWindows() with contextlib.suppress(cv2.error) for headless OpenCV builds