tool test: TestPython add more examples

[xushiwei]

• Proposal: MethodCheck — Logical-Type Method Detection in llcppg #945

[fennoai]

Review: generator test fixture for PyList_GetItem method generation

This PR extends the python-3.14.8 llcppg test fixture to exercise the C-global-function → Go-method path:

• include/pythread.h: adds PyAPI_FUNC(PyObject *) PyList_GetItem(PyObject *x, PyObject *index);
• llcppg.cfg: FuncPrefix ["Py_"] → ["Py_", "Py"]
• pythread.go (golden output): adds func (self *Object) Item(index *Object) *Object linked to C.PyList_GetItem

Conclusion: no blocking issues. The header input, config change, and golden Go output are mutually consistent.

Verified against the generator in cl/name.go / cl/func.go:

• FuncPrefix ordering ["Py_", "Py"] is correct and significant. rmPrefix returns on the first matching prefix, so Py_IsTrue → IsTrue and PyList_GetItem → List_GetItem (then → method Item via MethodCheck "Py*_: $1" + cutMethodPrefix). Reversing to ["Py", "Py_"] would mis-strip Py_IsTrue to a leading-underscore name. Existing golden entries (IsTrue, etc.) remain correct.
• The two-same-type guard in tryToMethod (cl/func.go) does not suppress this conversion. The guard is len(params) == 2 && params[1].Type() == t, which uses Go interface identity (==), not types.Identical. Both PyObject * params map to separate *types.Pointer allocations (types.NewPointer is not interned), so params[1].Type() == t is false, the guard does not fire, and the function is correctly emitted as the method (*Object).Item — matching the committed golden file. (Empirically confirmed with a standalone go/types check.)

Note (optional, non-blocking): include/pythread.h — the fixture declares PyList_GetItem(PyObject *x, PyObject *index), whereas the real CPython prototype is PyObject* PyList_GetItem(PyObject *list, Py_ssize_t index) (the index is an integer, not a PyObject*). This is harmless and intentional for a synthetic generation fixture, but since these files mirror CPython header layout, a brief comment noting the declaration is crafted to drive the method-generation test would keep a future contributor from "correcting" the type and silently changing coverage.

I could not run go test ./tool/ in this environment (the fixture test requires the llgo/libclang toolchain, which fails to link here), so the above rests on static analysis of the generator plus a direct go/types identity check rather than an executed golden-file diff.