Add capability for relationship weight to be a function of a concept value

[gregorbj]

It would be useful to make a relationship weight be a function of the value of a concept. This would be particularly valuable for quantity relationships (see issue #6). For example, say you have a model where the number of trips is a function of activity density (trips go up as density increases) and the trips are then split into non-auto and auto. The model fragment would be as follows:
Density ---> Trips ---> Non-Auto Trips
The first relationship would be a proportional one (the change in trips is proportional to the change in density). The second relationship would be a quantity one (the number of non-auto trips is a proportion of total trips). In strength of the relationship between non-auto trips and trips should change with activity density: as density increases the non-auto trip proportion increases. Therefore the relationship weight should be a function of density.

The model could be changed to recognize 7 types of weights.

1. Constant value
2. Function that increases at a constant rate as the concept value increases
3. Function that decreases at a constant rate as the concept value increases
4. Function that increases at an increasing rate as the concept value increases
5. Function that decreases as an increasing rate as the concept value increases
6. Function that increases at a decreasing rate as the concept value increases
7. Function that decreases at a decreasing rate as the concept value increases

The interface will allow the user to choose the type of weight. If a constant, then the interface would be similar to the present interface. If a function, the interface will allow the user to choose the type and also to set minimum and maximum weights.

All of the functional forms can be implemented using a logistic function, modifying the parameters, and scaling the results.