gh-92347: Throw a warning when converting a pointer to an iterator

[rruuaanng]

Similar to #92262. I think we should fix this (by setting tp_iter to None) but only on the main branch in case people rely on calling next() on a pointer object.

cc @JelleZijlstra

Since it is implemented in an extension, I added an exception when getting it. Do you think?

• Issue: Iterating over ctypes pointer causes hang or segmentation fault #92347

---

📚 Documentation preview 📚: https://cpython-previews--126318.org.readthedocs.build/

[rruuaanng]

It actually kind of overwhelmed me, the added exception also caused the CI to fail.

from test_ctypes/test_values.py

        class struct_frozen(Structure):
            _fields_ = [("name", c_char_p),
                        ("code", POINTER(c_ubyte)),
                        ("size", c_int),
                        ("is_package", c_int),
                        ]
        FrozenTable = POINTER(struct_frozen)

        modules = []
        for group in ["Bootstrap", "Stdlib", "Test"]:
            ft = FrozenTable.in_dll(pythonapi, f"_PyImport_Frozen{group}")
            # ft is a pointer to the struct_frozen entries:
            for entry in ft:
                # This is dangerous. We *can* iterate over a pointer, but
                # the loop will not terminate (maybe with an access
                # violation;-) because the pointer instance has no size.
                if entry.name is None:
                    break

Do note that indexed access to pointer objects is a documented (and useful) behavior. It would be quite surprising to have a sequence-like object raise an exception when used in a for loop, typing constructs notwithstanding.
Perhaps a note in the Pointers section mentioning the unbounded behavior when iterating aka an infinite generator.

Maybe we can just change the document.