KnSetWallBcs() is inaccurate for a moving obstacle

[zoharl3]

It leaves a dead zone near the ends of a moving obstacle. Instead, I implemented the algorithm in Bridson15, pg.72, fig. 5.2.

Also, I'm sure if setBound() or setBoundMAC() are correct for setting obvel, and I implemented my own version. (I didn't try to test it separately from the above to verify the culprit.)