Right, this is a bug in the VNL QuantumESPRESSO plugin: With nspins=2 (ordinary spin-polarized calculations), some charge-density data is written to "spin-polarization.dat". But with "noncolin=.true." the data is written to 3 files: "magnetization.x.dat", "magnetization.y.dat", and "magnetization.z.dat". VNL currently does not look for these files, but looks for spin-polarization.dat, which does not exist with noncolinear spin.
I have submitted a bug report on this, but I cannot say exactly when it will be fixed.