Fix lazy If branch pruning and skipped-parent handling in graph runtime

[JPPhoto]

Summary

Fixes lazy If execution in invokeai/app/services/shared/graph.py for workflows where a selected branch shares ancestors with the unselected branch.

The runtime bug was that resolving an If node decremented indegree for the pruned input but left the pruned execution edge in place, so later parent completion could decrement the same dependency again. This could make the If node become ready too early, underflow indegree, or propagate None into required downstream inputs.

This PR:

• prunes unselected If input edges from the execution graph during branch resolution
• ignores skipped prepared exec nodes when matching live parents for downstream materialization
• tolerates missing result payloads from skipped branch-local exec nodes during If input hydration
• adds focused backend regression tests for positive and negative branch-selection cases
• updates invokeai/app/services/shared/README.md to document the runtime behavior

Related Issues / Discussions

Complex workflow failure involving a single If node feeding positive_text_conditioning on flux_denoise.

QA Instructions

Load a workflow where collect -> if -> required downstream input is used and verify:

• the selected branch output reaches the downstream node
• branch-exclusive unselected ancestors are skipped
• no KeyError, indegree underflow, or missing required-connection error occurs from the If runtime path

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)