Hi I wanna ask. why does autocorrelation in estimated residuals at time t disappear when run in eviews but once the residuals are squared, it does not ?

what are conditional heteroskedasticity? is volatility clustering one of them? what about arch effects??