Translate np_where into the if else expressions

[gordonwatts]

The if then else expression should be directly supported by hep_tables.

Here is an example expression:

1.0 / dx if d.x != 0 else 0.0

This is implemented as

np.where(d.x != 0, 1.0/d.x, 0.0)

• [ ] Write a test similar to above
• [ ] Write a test where we take two constants (np.where(d.x < 0, -1.0, 1.0))
• [ ] Write a test where we do the same sequence twice: np.where(d.x < 0, np.sin(d.x), np.cos(d.x)
• [ ] Write a test with different leaves `np.where(d.x.pt < 0, d.x.eta, d.x.pt)
• [ ] Write a test where there is a bomb: np.where(d.x < 0, d.x, d.y)
• [ ] Update documentation.

It may be we can't do all of these given the way our multi-argument resolution works, but we should explore.

[gordonwatts]

We can't do anything interesting here without work on #33. I was hoping to avoid that.