RFCode centers of mass and covering the stimulus space

[zjroth]

Since we no longer use a fine grid to find regions in the stimulus space, we no longer have the resultant grid of codewords available for use elsewhere. This grid was used to ensure that we covered the entire stimulus space (when centers were generated randomly) and to approximate the centers of mass of the codeword regions. Without this grid, I do not know how to do either of these things. How do we fix these issues?

[zjroth]

I've added a grid to the constructor for RFCode so that we can ensure coverage. The initial implementation selects randomly from the uncovered points on the grid and then adds a center that is close (i.e., within half of the grid width) to the grid point; note that a point outside of $[0, 1]^2$ can be selected using this method (which might happen if one of the edge points of the grid is selected). An optional constructor parameter has also been added for control over the grid size (number of divisions in the grid).