Just something I noticed when reading through the code; I think the `io.Copy` was a remnant of an older version of that code. I don't see a reason why we don't just let it fall through and serve the file with `http.ServeContent` normally. More consistent that way.

Also, I hoisted the `status_code` override logic above the error handler detection bit, because then it allows for overriding the status code during error routes. Not necessarily that likely to be useful, but it's more correct. (Reminder that the first `w.WriteHeader()` wins, subsequent ones have no effect)

This makes sure additional header fields can be set by `http.ServeContent`, by only intercepting the `WriteHeader` call when `http.ServeContent` actually calls it.