Do we want `np.asarray(Array)` to work?

[nicola-bastianello]

is it even possible? this is of course not necessary as we expose to_numpy already, but it is very common to do such an operation

[Simpag]

It would be risky. We can make it work by adding __array__ method to the Array class but then we have to assume that every framework supports np.asarray(framework_tensor) and I think that might lead to bugs in the future. Furthermore, device handling might make it break

[nicola-bastianello]

good points, then let's only support to_numpy