Merge #41

41: Various improvements to FloatCore r=vks a=cuviper

- New macros simplify forwarding method implementations.
  - `Float` and `Real` use this to compact their implementations.
  - `FloatCore` now forwards `std` implementations when possible.
- `FloatCore` now requires `NumCast`, like `Float does.
- New additions to `FloatCore`:
  - Constants like `min_value()` -> `f64::MIN`
  - Rounding methods `floor`, `ceil`, `round`, `trunc`, `fract`
  - `integer_decode` matching `Float`'s
- Fix NAN sign handling in `FloatCore` (rust-num/num#312, rust-lang/rust#42425)
- Fix overflow in `FloatCore::powi` exponent negation.
- Add doctests to all `FloatCore` methods.

src/cast.rs

@@ -1,4 +1,3 @@
-use core::f64;
 use core::mem::size_of;
 use core::num::Wrapping;