Expose camera as translation vector + rotation quaternion

[Roland Groza]

Currently the camera is exposed as {lookAt, origin, up}, but this makes it difficult to work with when rotation operations are needed.

We should instead expose the camera as translation vector + rotation as a quaternion. As a consequence, rotation and translation operations can be achieved easier from the Python and/or JavaScript clients.

{
  "translation": [0, 0, 1],
  "rotation": [0, 1, 0, 3.14]
}

[Juan Hernando]

Can we close this one?