Fix dispatch of SparseMatrixCSC*Diagonal multiplication

[Pbellive]

This fixes #28934. In short, multiplication of a SparseMatrixCSC with a Diagonal matrix was getting dispatched to a generic matmul routine rather than the specialized methods defined at

julia/stdlib/SparseArrays/src/linalg.jl

Line 971 in c8450d8

function mul!(C::SparseMatrixCSC, A::SparseMatrixCSC, D::Diagonal{<:Vector})

because the type signature of the diagonal matrix input in those methods was wrong, as pointed out by @GiggleLiu.

Correctness of results is already tested at

julia/stdlib/SparseArrays/test/sparse.jl

Line 314 in 3fbdc9b

@testset "matrix multiplication" begin

I'm not sure how to write a test that checks that calling, say, A*D for compatible sparse and diagonal matrices dispatches to the right method. Is that doable/necessary?

[KristofferC]

Some sort of dispatch test could be done with e.g. (B is SparseMatrixCSC and D is Diagonal).

m = @which mul!(B, B, D);
@test m.module == SparseArrays

[Pbellive]

That dispatch test seems reasonable @KristofferC. Added dispatch tests to the PR.