Update CUSPARSE.sum_dim<n> to allow for arbitrary function on elements - #1987

julia> using CUDA, SparseArrays

julia> a = sprand(10, 10, 0.1);

julia> ad = CUDA.CUSPARSE.CuSparseMatrixCSR(a);

julia> sum(ad, dims=2)
┌ Warning: Performing scalar indexing on task Task (runnable) @0x00007f2882ffc010.
│ Invocation of getindex resulted in scalar indexing of a GPU array.
│ This is typically caused by calling an iterating implementation of a method.
│ Such implementations *do not* execute on the GPU, but very slowly on the CPU,
│ and therefore are only permitted from the REPL for prototyping purposes.
│ If you did intend to index this array, annotate the caller with @allowscalar.
└ @ GPUArraysCore ~/.julia/packages/GPUArraysCore/uOYfN/src/GPUArraysCore.jl:106
10×1 Matrix{Float64}:
 0.7312257183333077
 0.2736255562180383
 0.0
 0.8569093552724679
 0.24512118904635305
 0.0
 0.0
 0.0
 0.0
 0.9815900426493798

julia> using CUDA, SparseArrays

julia> A = sprand(4, 4, 0.8);

julia> dA = CUDA.CUSPARSE.CuSparseMatrixCSR(A);

julia> CUDA.CUSPARSE.sum_dim1(dA)
4-element CuArray{Float64, 1, CUDA.Mem.DeviceBuffer}:
 2.2032459284126857
 1.2487818875938437
 0.7308272178094658
 1.9313369732308794

julia> sum(A, dims=1)
1×4 Matrix{Float64}:
 1.40888  1.72323  0.602062  2.38003

julia> sum(A, dims=2)
4×1 Matrix{Float64}:
 2.2032459284126857
 1.2487818875938437
 0.7308272178094658
 1.9313369732308794