Why do a pair of variables with no significant correlation and no significant regression intercept and slope, have a highly significant regression with high adjusted $R^2$ when the regression is forced to pass through the origin?