Remove redundant conversion in structured matrix broadcasting (JuliaL…

…ang#55695)

The additional construction is unnecessary, as we are already
constructing a `Matrix`.
Performance:
```julia
julia> using LinearAlgebra

julia> U = UpperTriangular(rand(1000,1000));

julia> L = LowerTriangular(rand(1000,1000));

julia> @Btime $U .+ $L;
  1.956 ms (6 allocations: 15.26 MiB) # nightly
  1.421 ms (3 allocations: 7.63 MiB) # This PR
```

stdlib/LinearAlgebra/src/structuredbroadcast.jl

@@ -96,7 +96,7 @@ structured_broadcast_alloc(bc, ::Type{UnitLowerTriangular}, ::Type{ElType}, n) w
 structured_broadcast_alloc(bc, ::Type{UnitUpperTriangular}, ::Type{ElType}, n) where {ElType} =
     UnitUpperTriangular(Array{ElType}(undef, n, n))
 structured_broadcast_alloc(bc, ::Type{Matrix}, ::Type{ElType}, n) where {ElType} =
-    Matrix(Array{ElType}(undef, n, n))
+    Array{ElType}(undef, n, n)
 
 # A _very_ limited list of structure-preserving functions known at compile-time. This list is
 # derived from the formerly-implemented `broadcast` methods in 0.6. Note that this must