Add an ability to initialize Jupyter Visualizers such that starting camera is the original camera passed in

[shumash]

Current behavior: kaolin.visualize.IpyTurntableVisualizer and kaolin.visualize.FirstPersonVisualizer take an input camera and modify it in a difficult to predict way according to input arguments:

visualizer = kaolin.visualize.IpyTurntableVisualizer(
    cam.height, cam.width, copy.deepcopy(cam), render,
    max_fps=24, world_up_axis=2, focus_at=torch.zeros((3,), device=device))
# visualizer.cam is very different from cam

Desirable additional behavior:

visualizer = kaolin.visualize.IpyTurntableVisualizer(
    cam.height, cam.width, copy.deepcopy(cam), render,
    max_fps=24, focus_radius=1.0, device=device))
# visualizer.cam is the same as cam
# up axis and focus_at is determined from cam and focus_radius 

Why is this needed? Camera issues are prevalent in 3d research. It is very hard to debug cameras, if you can't set it to precise value.

[github-actions[bot]]

Stale issue, please reopen if still relevant