Rotational covariance for 3D poses not being displayed properly

[Ellon]

The current display on 126ba69 uses the same logic for positional and rotational covariances in 3D, which is not at all intuitive for the rotational covariance. A display like the one in the figure 2 of this article would be a better option.

[NikolausDemmel]

Just quoting myself on how the display could look for easier reference (quote from #10 (comment))

What I could imagine is changing the display of the orientation covariance to something that is similar to the linked-to paper above. I.e. one cone (or maybe just the 2d ellipse as in the paper) for each axis. For my 2D case that would degenerate to just two flat same-sized triangles along the x and y axes, where the length of the line that is perpenticular to the axis is 2sigma of the yaw variance ("radius" = 1 sigma). The cone along the z axis would not be visible (width 0). For me that would be nice to have, but it is not really vital currently.