Detect invalid polygons on PDF import

There seem to be a case where a pdf text extraction can generate an invalid polygon

See https://gitlab.com/arkindex/requests/-/issues/316