The doc for napi_define_class pretty much assumes an ObjectWrap scenario in that a C++ class will be associated one-to-one with the JS class being defined. This is not necessarily and always the case. We should limit the doc to what can be done in C.