`LineB*` over `A -> B` and `B -> A` do not always match

Same line segments in opposite octants can be off by one pixel. See (0,0)→(8,3) vs. (8,3)→(0,0):

```
FWD   BWD
(0,0) (8,3)
(1,0) (7,3)
(2,1) (6,2)
(3,1) (5,2)
(4,2) (4,1) <- off-by-one
(5,2) (3,1)
(6,2) (2,1)
(7,3) (1,0)
```

This is not a bug. Most implementations of Bresenham's algorithm are not symmetrical. However, if this were not an issue, implementing #24 would be easier.