Avoid ISO-8859-1 character in fr.po header. Python 3 gettext can handle arbitrary character sets for the messages in .mo files, but always tries to decode the header information as UTF-8. This means that using the French translation fails because it tries to decode b'Last-Translator: St\xe9phane Raimbault <stephane.raimbault@gmail.com>' as UTF-8, when that is an ISO-8859-1 string. Fixed by avoiding the accent in that string to make it pure ASCII. Alternatively, all the remaining non-UTF-8 translations might reasonably be converted fully to UTF-8.

--- a/CHANGES.txt	Thu Nov 21 20:50:56 2019 -0500
+++ b/CHANGES.txt	Fri Nov 22 14:31:16 2019 +0000
@@ -53,6 +53,8 @@
   when using validation. Added option to allow detection of duplicate
   username when the user tries to register. Previously user was
   rejected when confirming registration. (John Rouillard)
+- French translation gave errors with Python 3 because of ISO-8859-1
+  character in .mo file header. (Joseph Myers)
 
 2019-10-23 2.0.0 alpha 0