Sure, as long as you shift the voltage across the shunts to be in range of your ADC, e.g. centered around V
CC/2. But you lose the isolation from a hall effect sensor, so if this is critical (convenient for high-side sensing), then a hall sensor might be superior. The burden voltage will also be much higher.
Both shunts and hall sensors can deal with bidirectional currents, In the case of shunts you just have to prepare to handle both negative and positive voltages across the shunt. For Hall effect sensors you have to ensure that the one you pick is designed for bidirectional currents. For currents up to 35A the hall sensor seems to be the obvious solution to me, however, since you planned to have a shunt that can handle 30A, I just wanted to point out that you could do the same with just shunts.
If you are aiming for 0.2% accuracy, then I would say the ACS712 is already out with its 1.5% typical (not max) total output error. Unless you plan to calibrate and characterize the part yourself. And I am not convinced that the parallel resistor trick will do the accuracy any good. You would need to use a more accurate current sensor with ideally an accuracy < 0.1%. (to not eat up your entire error budget). Something like
this comes closer (but still only 0.25% accuracy.
You could go for a very low value shunt like
this one, and then have a precision or chopper op-amp amplify the +/- 70 mV full scale to the half the full scale of the ADC. This is not unlike the µcurrent. It might be good to read through Dave's
article and watch some of his videos about the µcurrent. You will have to figure out if this will provide sufficient resolution or if you need a second shunt like in the example I posted. This way you might just be able to meet that 0.2% spec, assuming you have a decent voltage reference for your ADC.
What is the lowest current you want to resolve?