Operations with Eye do not preserve sparsity

Hi, I've noticed that SquareEye, since it is a LinearAlgebra.Diagonal preserves sparsity in many operations. For example:

julia> using FillArrays

julia> using SparseArrays

julia> E = Eye(2)
2×2 Eye{Float64}

julia> M = sprand(2,2, 0.5)
2×2 SparseMatrixCSC{Float64, Int64} with 3 stored entries:
 0.906128   0.408236
 0.0285949   ⋅ 

julia> E + M
2×2 SparseMatrixCSC{Float64, Int64} with 4 stored entries:
 1.90613    0.408236
 0.0285949  1.0

julia> kron(E, E)  # although it would be nice is this were another SquareEye!
4×4 Diagonal{Float64, Vector{Float64}}:
 1.0   ⋅    ⋅    ⋅ 
  ⋅   1.0   ⋅    ⋅ 
  ⋅    ⋅   1.0   ⋅ 
  ⋅    ⋅    ⋅   1.0

However, general Eye typically converts to a dense matrix:

julia> E = Eye(2,2)
2×2 Eye{Float64}

julia> E + M
2×2 Matrix{Float64}:
 1.90613    0.408236
 0.0285949  1.0

julia> kron(E,E)
4×4 Matrix{Float64}:
 1.0  0.0  0.0  0.0
 0.0  1.0  0.0  0.0
 0.0  0.0  1.0  0.0
 0.0  0.0  0.0  1.0

This was quite surprising to me. Could we add some operator overloads to keep things sparse in the Eye case?

The issue names Eye, SquareEye, +, and kron but no files or tests; inspect the Eye and SquareEye definitions and existing operator methods first. Reproduce the supplied Julia examples, then add focused coverage showing that Eye operations retain sparse representations where appropriate and that the resulting behavior is consistent.