With regard to digitising and processing your data, how about using a standard PC soundcard in place of the Ardino? It'll obviously cover the full audio bandwidth (if that's all you're interested in) - I don't know what the samplerate available on an Ardino is, but I'd not be supprised if it was not that great. If you put (say) voltage signal into the left channel and current into the right, then you should get a minimal and very repeatable phase lag between current and voltage readings. You then have massive computing power at your disposal to exract whatever data you want, and you could even use Matlab or similar to minimise your programming time.
Thinking about it, I would be tempted to use transformers (the kind of little audio matching transformers for DI boxes etc) to obtain isolation. Use a simple potential-divider for the voltage and a small shunt resistor, with the transformer primary in parallel, to sniff the current. The transformer secondaries can be fed directly into the soundcard.
You'll have to think about how much burden you can afford to insert with your current shunt (it'll spoil the daming factor, of course). It may be possible to hack your amplifier to include it within the feedback loop to elliminate this effect.
It all depends really on how accurate and "instrumentation-like" you need the system to be; what bandwidth, how flat, how accurate, how noisy, how linear. A solution like I've just suggested would undoubtedly be cheap-and-cheerful, but might at least help you to work out how good a "propper" solution would need to be.