I'm somewhat surprized, that several contributors seriously suggested copper wire. But you have realized, why it isn't a reasonable option. For technical metering applications, 0.5 or 1% error over the specified temperature range won't be bad. I guess, you'll have difficulties to provide a better calibration for the instrument. But there's nothing against achieving a lower T.C. for the shunt of course. You also have to care for a suitable 4-terminal shunt circuit, not to lose accuracy due to unintended voltage drops.