Change GraphExecutionState's _get_next_node method to prioritize Iterate Nodes

[cgi-joe]

Summary

This PR optimizes the graph execution order by prioritizing Iterate nodes and their children, ensuring the correct execution sequence. The changes made to the _get_next_node method within the GraphExecutionState class address the issue of random output order and execution of Iterate node children.

Key changes include:

• Extracting Iterate nodes from the topological order and sorting them based on their index attribute.
• Prioritizing the execution of Iterate nodes and their children, ensuring the correct order.
• Falling back to the topological order if no Iterate node or its children are ready for execution.

These improvements provide a more predictable and intuitive behavior for Iterate nodes.

Related Issues / Discussions

QA Instructions

To test the changes in this PR:

1. Set up a workflow with one or more Iterate nodes.
2. Execute the workflow and inspect the output of the Iterate nodes (select the node in the workflow editor and inspecting the "Outputs" tab in the lower side panel)
3. Verify that the output order starts correctly with index 0 and that the item at index 0 is executed first.

Before this change, the index 0 item would be executed in a random order. With these changes, the execution order should be consistent and aligned with the index attribute of the Iterate nodes.

Merge Plan

Checklist

• The PR has a short but descriptive title, suitable for a changelog
• Tests added / updated (if applicable)
• Documentation added / updated (if applicable)

[psychedelicious]

Thanks for fixing this very long-standing and impactful issue!

Here's my test setup:

I also added a test to the PR for execution order.