Review the use of decode and encode of unicode strings

In general we cannot be sure that the stdout/stderr of the commands we call is valid unicode, so we have to protect all bytes.decode invocations with errors="surrogateescape", and make sure that all the matching str.encode also use errors="surrogateescape" (see https://docs.python.org/3/library/codecs.html#error-handlers)

• @clemenci started a discussion:

  As discussed on the chat, we cannot be sure that the output of our commands is valid unicode, so it's better to use:

              output["stdout"] = result.stdout.decode(errors="surrogateescape")
              output["stderr"] = result.stderr.decode(errors="surrogateescape")