Feature Request: Add feature gate implementing num-traits' Float, Signed and Unsigned traits for vectors

[soqb]

Motivation

• adding support for num-traits would decrease complexity and boilerplate for end-users integrating with libraries that depend on it:
  • num for abstract and generic number-work in rust
  • ordered_float for asserted not-Nan and orderable vectors
  • ndarray for higher-dimension arrays of vectors

Drawbacks

• reverts the changes made in Default impl of PartialOrd for VecN is a footgun #168 because PartialOrd must be implemented for types that are Num (minimal reference solution can be found here)
• increases complexity and compile times although locked behind a feature gate.
• some operations are nonsensical:
  • what does it mean to sin a Vec3?
  • is there any sensible way to convert a vector to a single element value?

Open Questions

• for those nonsensical cases described above we have a handful of options, which ones are applicable to which cases:
  • return None, Err or some insane default where applicable
  • find some sane fallback implementation, however unconventional
  • call unimplemented!()
• should this support be extended to matrices as well, or even just quaternions?

Additional Context

• W.I.P. reference implementation can be found on this fork - am happy to make a full PR if these changes are well-liked.