impl num_traits for Quantity.

[Tehforsch]

As discussed in #55 Quantity should probably implement at least some of the num_traits. In the current implementation, the methods are just anonymous methods that happen to be named the exact same way as the corresponding methods on some of the num traits.

For dimensionless quantities, virtually any traits that f32 fulfills should be possible on Quantity<f32, Dimension::none()> as well.
Mainly I am thinking of

• Num, NumAssign (and NumAssignOps), NumRef, NumAssignRef RefNum
• Float if we have either std or num-traits-libm activated, otherwise FloatCore
• Pow
• Inv

The following would be possible on all quantities:

• Signed
• Euclid

There are a few more and we'd have to look at what we need. For now, I'd probably start by moving the methods that were changed in #55 into the Float trait. Curiously, this trait isn't necessarily unit safe though, because of #57. I suppose we could overlook this for now, but I am not super happy about that.