`extract_text` returns bytes instead of string/unicode
See: https://framagit.org/kresusapp/kresus/issues/905
Should the extract_text
use StringIO, instead of a BytesIO here : https://git.weboob.org/weboob/weboob/blob/dfa8dc81dcb2b5c46cb7516acb1d45aa1e85b2b5/weboob/tools/pdf.py#L487