Can we use a smoother function for eta?

[c19]

In senarios inc happens not evenly in time, but a burst in certain periods, eta would jump high and low drastically. for example, _ means no inc, . means one inc. like this:

________________________________...__________________________..._______________.______________________...______

current eta seems like eta = elapsed_time/position*length-elapsed_time

A new update like new_eta = previous_eta * 0.75 + eta * 0.25 would make the eta smoother.

[djc]

See discussion in #580.