support unyt >= 3.0

[chrishavlin]

A couple of small failures related to the new unyt release, should be a quick fix to unpin unyt.

[chrishavlin]

e.g., the following code (in one of the tests, and in the readme) will fail due to the column stack of arrays with different units:

import matplotlib.pyplot as plt
# resample at a higher resolution
T = vbr.input.SV.T_K[:, 0, 0]
nT = len(T)
T_targets = np.linspace(T.min(), T.max(), nT * 2)
phival = vbr.input.SV.phi.min()
dgval = vbr.input.SV.dg_um.min()
phi_targets = np.full(T_targets.shape, np.log10(phival))
dg_targets = np.full(T_targets.shape, np.log10(dgval))
targets = np.column_stack((T_targets, phi_targets, dg_targets))
Vs_interp = interp(targets)

# compare to actual Vs along curve
actual_Vs = vbr.output.anelastic.andrade_psp.V[:, 0, 0, 0]

plt.plot(T_targets, Vs_interp, label="interpolated output", marker=".")
plt.plot(T, actual_Vs, ".k", label="actual VBRc output", markersize=12)
plt.legend()
plt.xlabel("T [K]")
plt.ylabel("Vs [m/s]")
plt.show()