Fix graph execution state resume after JSON round-trip

[JPPhoto]

Summary

Fix a graph execution resume bug in invokeai/app/services/shared/graph.py where GraphExecutionState could stall after model_dump_json() / validate_json() on a partially executed session. The fix rehydrates runtime scheduler state for already-materialized exec nodes after deserialization, and adds a regression test covering partial execution, JSON round-trip, and successful completion.

Related Issues / Discussions

QA Instructions

• Run pytest tests/test_graph_execution_state.py
• Confirm the new round-trip resume test passes
• Confirm a partially executed graph can be serialized, deserialized, and then finished without next() stalling
• Run ruff check --fix invokeai tests scripts
• Run ruff format invokeai tests scripts

Merge Plan

Checklist

• The PR has a short but descriptive title, suitable for a changelog
• Tests added / updated (if applicable)
• Changes to a redux slice have a corresponding migration
• Documentation added / updated (if applicable)
• Updated What's New copy (if doing a release after this PR)

[lstein]

Some suggestions:

1. _rehydrate_resolved_if_exec_branches duplicates scheduler logic. The condition-edge loop + copydeep at graph.py:1856-1861 duplicates code at _IfBranchScheduler._apply_condition_inputs (graph.py:177-188). I suggest refactoring so that the scheduling logic is at one place.
2. _rehydrate_prepared_exec_metadata sets source_node_id and state but iteration_path is left None. This is eventually fixed by _build_iteration_path, but it would be best to either immediately rehydrate iteration_path, or enter a comment that this is deferred to later.
3. Shouldn't there be a direct test for _rehydrate_resolved_if_exec_branches? I'd suggest adding a test where the graph contains an IfInvocation, execute it past the condition, round-trip the state, and assert the correct branch goes to completion.

[JPPhoto]

@lstein All should be addressed!

[lstein]

Looks good!