CSV imports don't handle different encodings or non-accented characters well

[sanjuroj]

Current behavior

CSV imports of files that contain non-ascii characters are not handled well. In some cases a UTF-8 encoded file with accented characters can't be edited. In other cases (e.g. "Comma Separated Values" or windows encoding from Mac version of Excel), the files won't be loaded at all.

Expected behavior

Handle at least UTF-8 ecoded documents well. It would be nice to detect Windows encodings too.

Steps to reproduce the behavior

For UTF-8 encoded file:

1. Save file from Excel with non-ascii characters using the "CSV UTF-8" format.
2. Upload to Tola Tables.
3. Click on a row to edit.
4. Try to save the changes. A UnicodeEncodeError is thrown with the message "'ascii' codec can't encode character u'\ufeff' in position 6: ordinal not in range(128)"

For other encodings, after the Table selection screen, an InvalidStringData error is thrown with the message "strings in documents must be valid UTF-8: 'test \x87ccented \x8e letters'"

Additional details