And, remember, because all the inputs are biased off the same source, they’ll all be at the same voltage. The implication of this: you only need one low-pass filter (probably on the voltage input) to derive the count that will remove the d.c. offset in all the other inputs to to that processor (because they all use the same ADC, it should be exactly the same value for each).
You don’t have to do this, emonLib will work fine anyway. (In fact, what you gain in samples per second probably won’t be worth the trouble.)