bug: pdf splitting modifies returned csv elements

Describe the bug
When specifying output_format as csv, the response from the api is different when split_pdf_page is True or False. When False, the elements contain an extra metadata field: text_as_html. This also means the element id does not match.

To Reproduce
_test_unstructured_client/integration/test_decorators.py::test_integration_split_csv_response illustrates this, but is passing because it asserts on a shortened string.

Expected behavior
The response to be identical whether or not split_pdf_page is True or False.

test_unstructured_client/integration/test_decorators.py::test_integration_split_csv_response から始め、短縮されたアサーションを、追加の text_as_html フィールドと異なる要素 id が明らかになる比較に置き換えます。split_pdf_page=True と False の CSV レスポンス処理を追跡します。両方のレスポンスが同一で、統合テストが完全な結果を検証すれば完了です。