refactor(extraction): deduplicate verbatim helper copies across language extractors

[ScriptedAlchemy]

Summary

Audit-driven refactor (item #2 of the latest structural health audit: the extractor duplication family). Deletes the helper functions that were copied verbatim across the ~25 src/extraction/*_extractor.rs modules and replaces them with shared implementations. No behavior change — bodies were moved unchanged, and extraction output was verified byte-identical (see test plan). Net −747 lines (766+/1513−) across 30 files.

One commit per helper family:

1. Docstring/comment helpers → new extraction::common
   • clean_comment: C-style variant (3 copies: c/go/glsl → clean_c_comment) and ///-aware variant (2 copies: cpp/objc → clean_c_doc_comment)
   • preceding-comment docstring loop (7 copies: c/cpp/go/objc/pascal/glsl + objc's extract_impl_method_docstring → docstring_from_preceding_comments, parameterized by the cleaner)
   • #-comment docstring loop (2 copies: bash/ruby → docstring_from_hash_comments)
2. extract_call_sites → common::extract_call_expression_sites (4 verbatim copies: c/cpp/glsl/hlsl; per-language fns remain as thin wrappers over their private ExtractionState)
3. Traversal helpers → existing extraction::traversal (which was added for exactly this migration and preserves search order)
   • find_child_by_kind (25 verbatim copies → find_direct_child_by_kind)
   • find_descendant_by_kind (4 verbatim copies: glsl/hlsl/objc/wgsl)
   • has_child_kind (2 verbatim copies: glsl/hlsl → has_direct_child_kind)
4. BASIC subroutine-synthesis helpers → new extraction::basic_common (gwbasic/msbasic2, cfg-gated): BasicLine, find_subroutine_ranges, derive_function_name, and the shared top-level-line filter (for_each_top_level_line)

Near-copies deliberately left unmerged

• rust_extractor::clean_comment — also strips //! inner doc comments (Rust-specific)
• pascal_extractor::clean_comment — handles { ... } and (* ... *) Pascal comments; its docstring fn now passes it into the shared loop
• powershell_extractor::find_descendant_by_kind — iterative and, unlike the deleted copies, matches the start node itself; not a verbatim copy
• typescript_extractor::has_child_kind — different semantics from the glsl/hlsl copies
• the remaining per-language extract_call_sites/extract_docstring variants — each keys on language-specific node kinds/fields; unifying them would need per-language parameterization that isn't a natural abstraction
• follow-up candidate (not in this PR): the per-extractor visit_children dispatch loops (~38 files) could route through traversal::visit_children, but that's a style migration rather than deleting verbatim helper copies

Each extractor keeps its own private ExtractionState, so shared helpers take the pieces they need (source bytes, file path, unresolved-ref sink) rather than the state struct; unifying ExtractionState itself is a separate, larger refactor.

Test plan

• Byte-identical output check: dumped normalized ExtractionResult JSON for all tests/fixtures/** files plus inline go/scala/hlsl/wgsl snippets (languages without fixtures) on the merge-base and after the refactor — diff clean
• cargo fmt --all --check
• cargo clippy --all-targets -- -D warnings
• cargo check --all-targets
• cargo nextest run -E 'binary(extraction_suite)' — 689/689 passed
• Full cargo nextest run --profile ci — 3341/3341 passed

[changeset-bot]

⚠️ No Changeset found

Latest commit: ce53aa0

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR