fix: report nested JavaScript function declarations once

[HarshRajSinghania]

Summary

• Nested JavaScript function declarations are now reported once by the complexity CFG builder.
• A declaration nested in another function declaration is kept, and a declaration nested in an arrow, method, or exported function is no longer emitted a second time as name_<line>.

Motivation
Fixes #135. Complexity analysis dropped inNamed-style declarations and duplicated inArrow / inMethod / inExported, which also inflated summary.total_functions.

Implementation

• processStatement now keeps CFGs built for declarations inside the function it just analyzed. Those nodes are already on the child CFG, so BuildAll will not rebuild them.
• When BuildAll sees a function span it already recorded, it stops the walk instead of descending. The AST walk visits the same statement from both Children and Body, and the previous return true walked into the body and minted the name_<line> duplicate.
• Nested function locations are marked after a function is built so a later walk of the same span is skipped.

Testing

• cd polyscan && go test ./internal/js/analyzer/ -count=1 — pass
• cd polyscan && go vet ./internal/js/analyzer/ — pass
• Added TestCFGBuilder_BuildAll_NestedFunctionDeclarations for the arrow, method, function-declaration, and exported-function cases from [BUG][auto] js/js-nested-function-declaration-dropped-or-duplicated: nested function declarations are reported twice (name, name_<line>) or not at all #135.

The ASCII-only name suffix path is unchanged for genuinely distinct functions that share a name.

[DaisukeYoda]

Thanks for picking up #135. markFunctionLocations fixes the duplicate on its own, but returning false from the BuildAll walk causes a regression.

Callbacks inside function bodies are dropped

The walk no longer enters function bodies, so function expressions inside them are never found:

function top(items) {
  return items.map((x) => { if (x) { return 1; } return 0; });
}

main reports anonymous_2, but this branch does not. Returning true in both places fixes it. With that change, #135 stays fixed and go test ./internal/js/... passes. Please also add a callback case to the new test.

Minor (optional)

When two functions each contain a nested function h(), absorbNestedFunctionCFGs drops the second h. This is not a regression, so a follow-up PR is fine.

[HarshRajSinghania]

Thanks for the review @DaisukeYoda.

Addressed the regression: the BuildAll walk now returns true in both places so it continues into function bodies and discovers nested callbacks / function expressions. markFunctionLocations still prevents the name_<line> duplicates.

Also added the callback case from your example to TestCFGBuilder_BuildAll_NestedFunctionDeclarations.

Verified with go test ./internal/js/analyzer/ -count=1 (pass).

The optional absorbNestedFunctionCFGs name-collision note is left for a follow-up as suggested.

[DaisukeYoda]

LGTM