The magnetic moment of the electron is a magnetic moment, so the right magnetic field around it is
$$ \mathbf{B}({\mathbf{r}})=\nabla\times{\mathbf{A}}=\frac{\mu_{0}}{4\pi}\left(\frac{3\mathbf{r}(\mathbf{\mu}\cdot\mathbf{r})}{r^{5}}-\frac{{\mathbf{\mu}}}{r^{3}}\right). $$
The world is quantum mechanical – and so is any viable description of the spin – so we have to respect the postulates of quantum mechanics. In particular, the magnetic field above corresponds to a "state" (e.g. spin up and spin down) and one may construct complex linear superpositions of such states. It is important to realize that the quantum mechanical superposition of states in the Hilbert space in no way implies that the corresponding magnetic fields are being added according to the classical electromagnetic field's superposition principle.

Indeed, they're linear superpositions of states that contain different profiles of the magnetic field.

The magnetic field around the electron is so weak that indeed, it can in no way be thought of as a classical magnetic field, in the sense that the classical fields are "large". But the classical *formula* for the magnetic field is still right! This formula *defines* the magnetic moment. The quantum effects are always important, however. Also, if you try to measure this very weak magnetic field, it will unavoidably influence the state of the measured system, including the electron's spin itself.

It is of course completely wrong to imagine that we could measure such a weak magnetic field by a big macroscopic apparatus, like a fridge magnet. The effect of one electron's magnetic field on such a big object would be nearly zero, of course. In fact, quantum mechanics guarantees quantization of many "phenomena", so instead of predicting a very tiny effect on the fridge magnet, it predicts a finite effect on the fridge magnet that occurs with a tiny probability.

You may "measure" the electron's magnetic field by creating a bound state with another magnet in the form of an elementary particle. For example, the electron and the proton in a hydrogen atom are exerting the same kind of force that you would expect from the usual "classical" formulae – but it's important to realize that all the quantities in the equations are operators with hats.

Let me show you an example of a simple consistency check implying that there is no contradiction. Calculate the expectation value of the magnetic field (an operator) $\vec B(\vec r)$ at some point for the state $c_{up}|up\rangle + c_{down} |down\rangle$. The column vector of the amplitudes $c$ is normalized. The check is that you get the same expectation value of $\vec B(\vec r)$ for each $\vec r$ if you first compute it for the "up" component and "down" component separately, and then you add the terms, or if you first realize that it's a spin state "up" with respect to a new axis $\vec n$, and compute $\vec B$ from that.

It's a nice exercise. The point is that the expectation value of $\vec B(\vec r)$ is a bilinear expression in the bra-vector and ket-vector $|\psi\rangle$, much like the direction $\vec n$. And indeed, $\vec B$ is linear in the direction $\vec n$, according to the formula above, so things will agree. You may insert anything else to the expectation value, in fact, so the check works for all linear expressions in $\vec B$. The higher powers of $\vec B$ also have expectation values but they will behave differently than in classical physics because there will be extra contributions from the "uncertainty principle", analogous to zero-point energies of the harmonic oscillator in quantum mechanics.