A differential amplifier would probably be enough if only low voltage "isolation" is needed. If high voltage, high frequency isolation is needed, I have thought about modulating a RF carrier with the signal, isolating it using a transformer, and then doing synchronous detection.
You don't need isolation for high voltage, just enough dynamic range (or attenuation to bring it into range), plus safety measures that limit the fault current to a safe level. This is how the commercial high voltage differential probes do it (of course high voltage in this context is anything above ELV, not what power guys call high voltage). It's pretty hard to have attenuation that won't destroy your CMRR, however, since for even 100:1 CMRR (which is pretty bad), the attenuation of both inputs needs to be within 1% over the frequency range.
If you do need isolation, you're definitely talking about something much harder. Modulation would probably be the way to go (that's how some commercial designs do it, although some combine it with optical for low-frequency part), but orders of magnitude more complex than the uCurrent. Tektronix did in their TPS2000 series scopes, and I've read a fair amount of complaints about noise and slew rate.
There's a Tektronix service manual for their isolation amplifier with full schematics floating around (older than the TPS2000 series, back when they still published schematics), although it's not exactly a design you'd replicate on a breadboard.