fix(workflow): preserve conversation array appends across Iteration rounds

[Dexterity104]

Fixes #34417

Summary

Sequential Iteration runs a separate child graph engine per item with a deep-copied variable pool. Variable Assigner updates (e.g. appending to a conversation array[...]) were still applied to the parent pool only when the parent engine's dispatcher handled NodeRunVariableUpdatedEvent. That could happen after the next iteration had already taken another deep copy of the parent pool, so one append could be missing, especially visible on a later chat turn when the array was non-empty.

This change adds an IterationVariableSyncLayer on child engines that use an isolated pool (the Iteration pattern). On each NodeRunVariableUpdatedEvent, the layer writes the same variable into the parent pool inside the child's EventManager.collect path, so the parent pool is up to date before the child run finishes and the next iteration copies it. Loop child engines are unchanged (they share the parent pool and never pass a custom variable_pool).

• New layer: api/core/app/workflow/layers/iteration_variable_sync.py
• Wired in: api/core/workflow/workflow_entry.py (_WorkflowChildEngineBuilder.build_child_engine)
• Tests: api/tests/unit_tests/core/workflow/test_iteration_variable_sync_patch.py and updated expectations in test_workflow_entry_helpers.py

Checklist

• This change requires a documentation update, included: Dify Document
• Understand that this PR may be closed in case there was no previous discussion or issues. (This doesn't apply to typos!)
• Added tests for the change; kept the diff focused on this fix.
• Updated product documentation (not applicable for this internal engine fix unless maintainers want a note in workflow docs).
• Ran backend checks: uv run --project api pytest api/tests/unit_tests/core/workflow/ (and targeted tests for the new layer); run make lint / make type-check from api before merge if CI differs from local setup.

[Dexterity104]

Hi @QuantumGhost @laipz8200 @Yeuoly @crazywoola
Can you please review this PR when you have some time?

[Dexterity104]

Hi @QuantumGhost @laipz8200 @Yeuoly @crazywoola

[Dexterity104]

Hi @asukaminato0721 ,
I'm sorry to bother you. I just wanted to ask if there's a reason my PRs have been taking longer to review. I've noticed that PRs from @YB0y are being merged quite quickly, so I wanted to check if there's anything I should improve or adjust on my side.