aleviating wilson's frustration
if you look at wilson's algorithm, as presented here (and coloured here) by mike bostock, you might notice that before the first few paths are laid down, it can be painful to watch the random walk try to find the existing maze.
here, i modified his algorithm by adding walls, instead of cutting paths. it has the advantage of starting with the entire edge as a target, so it goes much faster at the start.
i also made each random walk start at a random point, rather than iterating through in row-major order.