Pixel density refers to how small the texture pixels are on the model surface. The less unused UV space, and the larger the bitmap, the smaller those pixels will be, giving you greater detail to work with.
In both links i find a term 'pixel density' used but not exactly explained. They both suggest having more dense uv's for face and upper body, which makes sense. But I am not sure if I'm right in thinking that it would actually be much harder to texture face with many polygons and retain desired amount of detail if the uv's…