TEST Fix flaky XPIA unit tests by stubbing workflow._memory

[hannahwestra25]

Problem

Four tests in test_xpia.py were failing in CI with RuntimeError: memory offline:

• test_perform_async_complete_workflow_with_scorer
• test_perform_async_workflow_without_scorer
• test_perform_async_scorer_returns_empty_list
• test_perform_async_scorer_raises_exception

Root cause

These tests exercise XPIAWorkflow._perform_async, which calls _execute_processing_async, which in turn writes the processing response to memory via self._memory.add_message_to_memory(...). Unlike test_execute_processing_async_adds_to_memory, the failing tests did not replace workflow._memory with a mock, so they depended on a live central memory backend that isn't reliably available in CI.

Fix

Update the shared workflow pytest fixture (and the inline workflow construction in test_perform_async_workflow_without_scorer) to stub workflow._memory with a MagicMock. This keeps the tests properly unit-scoped and isolates them from the real memory backend without changing production behavior.

Tests and Documentation

Ran updated tests