Missing dot product specializations for `SparseArray`s with `Adjoint{<:SparseArray}`s

[MasonProtter]

Here's an MWE where there's a 100x performance drop due to hitting the generic dot fallback when taking the inner product of a sparse array with the adjoint of a sparse array:

julia> using SparseArrays, LinearAlgebra, BenchmarkTools

julia> let N = 1000
           A = sprand(N, N, 0.1)
           At = spzeros(N, N)
           adjoint!(At, A)
           @btime $At ⋅ $A
           @btime $A' ⋅ $A
       end;
  509.345 μs (0 allocations: 0 bytes)
  32.945 ms (0 allocations: 0 bytes)