Describe the feature you'd like
If an example prints arbitrary data, there is no way to avoid a unicode encoding/decoding error.

Instead of failing, byexample could replace the offending bytes with something else (even delete them). This strategy could be set from the same --encoding parameter.

See in Python 3 help(str.encode) and the man page of Python about PYTHONIOENCODING.

Make the change backward-compatible.

Additional context (optional)
Is strict, the current strategy, the best as default? We could change it too to backslashreplace
If we choose this last one, we should add a note or warning if we perform the replace op following the line of #77 (how to detect the replacement, no idea!)