Change type signature for Base.:+ square args

[mtsokol]

Hi @willow-ahrens,

I noticed that for:

using Finch
using LinearAlgebra

A_ref = [1 2 0 4 0; 0 -2 1 0 1]
A = Tensor(Dense(SparseList(Element(0))), A_ref)
norm(A, 1) == norm(A_ref, 1)
norm(A, 2) == norm(A_ref, 2)  # fails
norm(A, 3) == norm(A_ref, 3)

the square option fails with an error:

ERROR: MethodError: no method matching +(::Finch.Square{Int64, Float64}, ::Finch.Square{Float64, Float64})

Closest candidates are:
  +(::Any, ::Any, ::Any, ::Any...)
   @ Base operators.jl:587
  +(::T, ::T) where T<:Finch.Square
   @ Finch ~/Finch.jl/src/interface/lazy.jl:424

This PR fixes it.

[mtsokol]

Ok, looking at the CI I don't think my fix is the right one. But the issue described here is still relevant for computing vector norms with p=2.