Enable operations on state variables before the neural activity is projected onto a monitor's space

This issue concerns models and monitors and how their are linked through
variables of interest.

#Background:
The relevant input neural activity for the different types of biophysical
recordings sometimes corresponds directly to one of the model's state
variables. However very often we found that the what is measured is the result
of the difference or summation of two or more state variables.

#Current implementation:
The recently introduced Epileptor is one example. The Jansen-Rit and
Zetterberg-Jansen models are good examples as well. The latter has an extra
state variable to describe the difference. However, this remains inside the
model. With the current implementation, if we wanted to see difference between
another pair of state variables, we should define yet another equation.

#What we need:
A mechanism specific to the monitors that performs a weighted sum with the set
of state variables.

Example:
Let x and y be the variables that describe the temporal evolution of a certain
quantity 'q', for two different assemblies of neurons (ie, two neural masses).
The input neural activity required for computing the BOLD signal is, for
instance, x-y.

At the implementation level, this would have x and y as the variables of
interest and the weights should be w_x = 1 and w_y= -1.

After performing the operation the neural activity is projected onto the
corresponding biophysical state.