Describe the feature you'd like
There are cases where an example may output binary data or just garbage. Even if the example uses <...> to ignore it, the underlying runner (based on pexpect) will complain of an unknown encoding and in this case there is no valid encoding to save this issue.

The only alternative is to tell pexpect that ignore or replace the invalid code points.

Python supports this and for the PYTHONENCODING variable, Python uses the syntax encoding[:errorhandling] and byexample could use the same for its --encoding parameter.