UniformScaling: allow writing generic polynomicals with I.

[StefanKarpinski]

julia> p(x) = x^2 - 2x + I
p (generic function with 1 method)

julia> p(4)
9

julia> p(rand(3,3))
3x3 Array{Float64,2}:
  0.386215   -0.22931    0.0859286
 -0.18856     0.664356  -0.693054
  0.0600554  -0.361883   0.619314

[StefanKarpinski]

cc: @JeffBezanson, @andreasnoackjensen, @jiahao, @stevengj, @ViralBShah

[JeffBezanson]

No objections here.

[toivoh]

+1 for UniformScaling + Number = Number, which I take it is the heart of this. (LGTM)

[StefanKarpinski]

Yep, that was the change. It's an interesting case because how you "promote" is different for different for different arithmetic operations.

[andreasnoack]

👍

[jiahao]

Also cute:

julia> p(I)
UniformScaling{Int64}(0)

[StefanKarpinski]

That is cute.

[toivoh]

It's also a nice sign of consistency. Thanks for bringing it up!

[carlobaldassi]

Looks like the scope of this type could (should?) be extended to any ring, becoming part of the (informally defined) interface provided by zero, one, +, *. May make sense to call it ScaledUnity at that point?
It also occurs to me that we should define one(::UniformScaling) and companions, in light of that.

[carlobaldassi]

(Update: I just added one and zero to UniformScaling, mostly because why not?)