experiment: remove shouldSkipCurrentNode guard

[TomasVotruba]

Experiment: what breaks if we remove the shouldSkipCurrentNode guard?

Removed from RectorRunner::run():

// node already changed by this rule in a previous pass -> hard skip
if ($this->skipper->shouldSkipCurrentNode($rector::class, $node)) {
    return null;
}

Result

ChangeSwitchTernaryTest fails:

ShouldNotHappenException: Scope not available on "PhpParser\Node\Stmt\Expression" node.
Fix scope refresh on changed nodes first   (ScopeFetcher.php:26)

Why

The guard delegates to RectifiedAnalyzer::hasRectified(), which covers two cases:

1. consecutive CREATED_BY_RULE - same rule re-applies to a node it just created (infinite re-apply loop).
2. just-reprinted node with overlapping token start and no refreshed scope.

Case 2 is the one that breaks. When a rule reprints a node (switch -> match here), the new Expression node has no scope yet. The guard skips it before refactor(). Without the guard, a rule subscribing to that node runs, calls ScopeFetcher::fetch() on a scopeless node, and throws.

tests/Issues/ChangeSwitchTernary exists precisely to pin this behavior.

Alternatives to handle it differently

• Refresh scope on reprinted nodes before running rules (so ScopeFetcher never sees a scopeless node) - heavier, defeats the skip's purpose.
• Split the two concerns into separate explicit checks - clarity only, same work.

Current guard is the cheapest correct approach. Draft - not for merge, demonstrates the finding.