Which knots are singularities of a hyperbolic cone-manifold structures on $S^3$?

The results proved in

S. Kojima, "Deformations of hyperbolic 3-cone-manifolds", J. Differential Geom. 49 (1998), no. 3, 469-516

provide complete answers to questions 1 and 3.

The main theorem of the cited paper ensures that any compact hyperbolic cone metric with cone angles at most π can be continuously deformed to a complete hyperbolic metric on the complement of the singularity. Furthermore this deformation can be done by an angle decreasing deformation.

This provides a partial answer to your question (2) and a complete answer to (3): If $K$ is the singular locus of a cone structure with angles $\leq \pi$, then $K$ is necessarily hyperbolic. Moreover, thanks e.g. to a suitable version of the Schlafli formula, the hyperbolic volume of the cone structure is stricly smaller than the hyperbolic volume of the cusped manifold $S^3\setminus K$, which in turn is equal to the simplicial volume of $S^3\setminus K$ (up to the constant $v_3$, which I guess you forgot to mention when asking question 3).

Moreover, Theorem 1.2.1 of Kojima's paper implies that, if $S^3$ supports a cone structure with singularity $K$, then $K$ must be hyperbolic, regardless of the cone angle of the original cone structure along $K$.

Let me add some remarks in addition to Roberto's answer.

Question 2. has the same answer as Question 1: a knot is the singular locus of a hyperbolic cone metric iff the knot has hyperbolic complement (complete finite-volume hyperbolic metric) iff there is a cone metric with angle $\leq \pi$. This follows from Thurston's hyperbolic Dehn filling theorem: the complete metric may be thought of as a metric with cone angle zero, and the angle may be perturbed to be $\epsilon < \pi$.

For 3., the volume relation holds whether or not the cone angles are $\leq \pi$. When the angles are $\leq \pi$, Kojima shows that the cone angle may be continuously deformed to $0$, and that the volume increases during this deformation by Schlafli's formula. As far as I know, an analogous deformation result is still unknown for cone angles $\geq \pi$. However, there is a global comparison result which implies that the volume is at most the simplicial volume of the knot complement. This follows from the "natural map" technique of Besson-Courtois-Gallot. One may also prove this using Gromov's approach of measurable cycles and simplicial volume (Ben Klaff did this in his thesis).