fix XmlElement::pop_element

[DenSinH]

See the test.

Basically, what was happening before was the following:
If you define a model with ElementT fields, the deserializer would deserialize the field as a verbatim copy of the element, but it would not pop it from the parent element. If your model has extra="forbid", the ModelSerializer would then check for any extra fields, and it would find all of the ElementT fields, as they have not been popped.

In the old situation, pop_element would practically be the same as find_element, not modifying the element, but even judging from the name, this is a bug. The test shows the desired behavior, where extra="forbid" would not mark any ElementT elements as being extra.

[DenSinH]

Okay apparently it is intended behavior to not actually pop elements in pop_element, I am not sure what you would prefer, a change in the signature for pop_element, a new method pop_and_remove_element or perhaps some private-member access in raw.ElementSerializer::deserialize like so:

    def deserialize(
            self,
            element: Optional[XmlElementReader],
            *,
            context: Optional[Dict[str, Any]],
            sourcemap: Dict[Location, int],
            loc: Location,
    ) -> Optional[str]:
        if self._computed:
            return None

        if element is None:
            return None

        if (sub_element := element.pop_element(self._element_name, self._search_mode)) is not None:
            # private member accesses:
            element._state.next_element_idx -= 1
            element._state.elements.pop(self._state.next_element_idx)
            sourcemap[loc] = sub_element.get_sourceline()
            return sub_element.to_native()
        else:
            return None

It still seems to me like it is in fact a bug that forbid="extra" and native.ElementT interact in the way they do in the current version, but one of these would be the best fix.

[dapper91]

@DenSinH thanks for the report.

Fixed the bug in the PR #240