Specialize newindex for StructuredMatrix broadcasting (#55626)

This provides most of the benefits seen in
#55604. The simpler
implementation appears to help with branch-prediction in inferring the
zero elements of the structured matrices.
The improved performance as a consequence:
```julia
julia> using LinearAlgebra

julia> U = UpperTriangular(rand(3000,3000)); D = Diagonal(rand(size(U,1)));

julia> @Btime $U .+ $D;
  23.405 ms (3 allocations: 68.66 MiB) # nightly
  15.266 ms (3 allocations: 68.66 MiB) # This PR
```

---------

Co-authored-by: Matt Bauman <[email protected]>

Author: jishnub

stdlib/LinearAlgebra/src/structuredbroadcast.jl

@@ -199,6 +199,8 @@ function Broadcast.newindex(A::StructuredMatrix, b::BandIndex)
     # and we apply newindex to both the axes at once to obtain the result
     size(A,1) > 1 ? b : BandIndex(0, 1)
 end
+# All structured matrices are square, and therefore they only broadcast out if they are size (1, 1)
+Broadcast.newindex(D::StructuredMatrix, I::CartesianIndex{2}) = size(D) == (1,1) ? CartesianIndex(1,1) : I
 
 function copyto!(dest::Diagonal, bc::Broadcasted{<:StructuredMatrixStyle})
     isvalidstructbc(dest, bc) || return copyto!(dest, convert(Broadcasted{Nothing}, bc))