Updated sensor models

[motters]

A couple of images showing sensor models for a large and small visual
disparity.

http://code.google.com/p/sentience/wiki/StereoSensorModel

Previously I just used a simple stand-in function for stereo rays,
which I knew was definitely wrong but gave me a quick and dirty way to
do some grid mapping experiments. The new code for calculating sensor
models should be much more mathematically accurate. Hopefully this
will give the best possible chance for the system to be able to
resolve small objects in the presence of significant uncertainty.