Correct implementation of the Reynolds model

[mearlboro]

Currently the model's behaviour is unrealistic, and it's unclear whether it is because of implementation or parameters or boundary conditions.

In particular, boids seem to have a tendency to turn clockwise and spin in place like a bunch of dizzy drunkards.

There also seems to be a bug with plotting the trajectories.

Trajectories for the two cases above are in reynolds-data.zip

[mearlboro]

The jump from $\pi$ to 0 was a bug in angle_mod and has been fixed in 6d4eb97

[mearlboro]

Likely the implementation of the model is faulty esp in bearing_to which is used by the two non-Vicsek rules, avoidance and aggregate.