NumPy float arange() example no longer illustrates the problem

Our example is:

```pycon
>>> numpy.arange(0.0, 50.0, 0.3)[-1]
49.799999999999997
```

But this gives the following with current Python and NumPy:

```pycon
>>> numpy.arange(0.0, 50.0, 0.3)[-1]
49.8
```