fix(extract): emit references edges for Python decorators (#2154)

[Rishet11]

Closes #2154

Python now emits the same decorator edges TS/JS already emits.

Problem

Applying a Python decorator emitted no edge to the decorator symbol, so affected <decorator> answered "No affected nodes found" for every function it wraps — a silent false negative on reverse-impact queries.

decorated_definition was handled as a transparent wrapper so a decorated method keeps its class-qualified node id (#1050). Its decorator children were walked, but no branch consumed them:

if t == "decorated_definition":
    for child in node.children:
        walk(child, parent_class_nid=parent_class_nid)
    return

TS/JS already handled this via _ts_emit_decorator_edges. Python was never wired up.

Fix

A _python_decorator_name helper placed directly above its twin _ts_decorator_name, differing only in grammar node names (Python call/attribute vs TS call_expression/member_expression), plus emission inside the existing decorated_definition branch.

• Same shape as TS: a references edge with context="decorator" from the decorated function/class to each decorator symbol. No new edge type, so no schema, HTML exporter, MCP server, or affected change.
• Owner ids reuse the definition branches' own formulas verbatim (_make_id(stem, ns, name) for classes, _make_id(parent_class_nid, name) for methods, _make_id(stem, name) for module-level functions), so the edge lands on the node the walk is about to create.
• Targets go through ensure_named_node, so an imported decorator becomes a sourceless stub the corpus rewire collapses onto its real definition — this is what makes cross-file affected work.
• Gated on config.ts_module == "tree_sitter_python"; decorated_definition is a Python-only grammar node.
• The #1050 parent_class_nid propagation is untouched, with a regression test pinning it.

Covered forms: @deco, @deco(arg), @mod.deco(arg) (targets the symbol, not the module alias, matching _ts_decorator_name), stacked decorators, decorated classes, and decorated methods. Names that normalize_id() empties are skipped, matching the #1899 guard in the definition branches, so no edge can dangle.

Verification

End-to-end on the issue's repro, before:

Affected nodes for my_decorator()
- consumer.py [imports] pkg/consumer.py:L1

after:

Affected nodes for my_decorator()
- consumer.py [imports] pkg/consumer.py:L1
- business_logic() [references] pkg/consumer.py:L4

Check	Result
New tests before the fix	7 of 9 fail (real negative controls)
New tests after	9 passed
tests/test_ts_decorators.py + Python extraction subsets	61 passed
Full suite, base 66d8110	15 failed, 3620 passed
Full suite, this branch	15 failed, 3629 passed
Failure-set diff	identical, no new failures
ruff check	passed
python -m tools.skillgen --check	OK, 134 artifacts match

The 15 pre-existing failures are unrelated (test_ollama.py backend-detection and similar) and identical before and after; measured with -p no:randomly in the same env.

Known limitation

A comment placed between @ and the decorator expression (@ # note newline deco) drops that one edge. It is a missed edge, never an incorrect one, and does not affect decorators written in normal style. Left out of scope to keep the diff surgical; happy to fold in a fix if you would prefer it here.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.