constructors for Matrix and SparseMatrixCSC from UniformScaling

[Sacha0]

This pull request provides constructors for Matrix and SparseMatrixCSC from UniformScalings, for example Matrix(I, dims) and SparseMatrixCSC(I, dims). In other words, this pull request explores the intersection of #11557 (particularly #11557 (comment)) and JuliaLang/LinearAlgebra.jl#454. So far this approach feels nice, supporting the direction in #11557 (comment). Best!

[Sacha0]

Ref. JuliaLang/LinearAlgebra.jl#454, suggesting special-casing Matrix(I, m, n)/sparse(I, m, n)/SparseMatrixCSC(I, m, n) to yield an array with eltype Float64 rather than eltype(I).

[andreasnoack]

As mentioned in JuliaLang/LinearAlgebra.jl#454, I don't think it is necessary to return Float64 elements. Using Int elements will probably also result in a better promotion in some cases. At least for Float32 computations.

[Sacha0]

Cleaned up the existing commits and added a couple commits providing the convenience forms suggested by @andreasnoack's JuliaLang/LinearAlgebra.jl#454 (e.g. Matrix[{T}](I, n), SparseMatrixCSC[{T}](I, n)).

This pull request should be in shape. Absent objections or requests for time, I am inclined to move forward with this approach (merge) in the next day or two. Thoughts? Thanks!

[Sacha0]

Thanks all!