Even though CSV encoding is not specified, the de-facto "standard" nowadays seems to be utf-8.

• Google spreadsheet, export in utf8 by default and autodetect at install
• libreoffice ask users when saving and importing
• with Excel, users can choose from the list of formats: "CSV" (which is latin1, at least, with a french locale ) or "CSV UTF-8' (which is utf-8)

The problem was that cloudooo only accept latin1 and because latin1 is a subset of utf-8, it's not even possible to workaround before sending to cloudoo.

So this MR is about changing default encoding for cloudooo importing CSV from latin1 to UTF-8. Some backward compatibility is kept, if file cannot be decoded as utf-8, we just use latin1.

There are some other slightly unrelated changes, but since we are here:

• "modernize" tests a bit, by using unittest builtin loader instead of re-implementing something. Also split a big class in smaller classes to group tests by topic.
• fix some warnings at startup about deprecated arguments
• cloudooo also only supported ; as field delimiter. This is now more clever thanks to python's csv.Sniffer
• Correct current failures from master. I did not try too much, but ignore docy the same way doc was already ignored. We can do better if you want.