Transport coefficients in new cantera are wrong

Description of the problem

Transport coefficients (viscosity, thermal conductivity) calculated via PoKiTT are wrong. I've tracked this down to the polyfit function inside cantera where the species viscosities and thermal conductivities are fit to polynomials. The inputs to that function are all consistent, but the function returns different polynomial coefficients in Zodiac and PoKiTT.

Strangely, PoKiTT doesn't show this problem. It gets the expected result.

I suspect that other downstream usage (Wasatch, ODT) will have the same problem.

Note that the polyfit function calls through to the Eigen library, which Cantera builds internally. SpatialOps also builds Eigen internally. However, since it is a header-only library, I don't see how we could get a clash there...