in 5ae0508d (accounting_l10n_fr: Workaround encoding bugs
in Test Compta Demat, 2023-03-31) we replaced a few characters
that were known to be problematic, but while trying on some
other production data, we found new problematic characters.
This shown that this approach was too "optimistic" and that we
the safest way to have a file that is compatible with Test
Compta Demat is to use only ascii characters, this use
unicodedata to try to retain the original characters (so that
"Jérôme" becomes "Jerome") and also special case € to replace
it by "EUR"