Reconstruct full 3D fields from cylindrical simulation

[RemiLehe]

This allows openPMD-viewer to return a 3D cartesian array, from a file contains the thetaMode representation of the fields.
More specifically, ts.get_field will return a numpy 3darray (and corresponding info object) if the user passes theta=None (note that the default value for theta is 0)

This feature can be useful for postprocessing operations that are easier on a 3D cartesian grid (e.g. computing the total energy). One of the motivations of this PR is also to allow VisualPIC to work with cylindrical data (including 3D visualization of the fields).

Regarding the implementation:

• I updated the functions that read the field in quasi-3D, to support theta=None.
• I added an example in the tutorial notebook.
• Note that right now, the function construct_3d_from_circ typically dominates the time to get the field. I tried to speed up this function using Cython (see this commit), but this made it 10x slower. I obviously did something wrong but I am not sure what. So in the end, this function remained in Python/Numpy in this PR.

[RemiLehe]

The code below is unchanged (just reindented)

[soerenjalas]

Wow this is pretty cool. In terms of API, I don't have a strong opinion. I think the theta=None option works. Even though, this in a way then might seem that this is the default behaviour.

[RemiLehe]

Thanks for the feedback.
In fact this is doing nearest-grid-point interpolation (so not really an interpolation, as you noted). But I'll maybe replace it by linear interpolation in a separate PR.