Basic arithmetic operation fails on `Tensor`s with matching `inds` but different permutation

[jofrevalles]

When attempting to perform a sum or subtraction between two tensors that have matching inds but in a different permutation, the operation fails with a DimensionMismatch error. I instead expected that these operations were possible.

Example:

julia> using Tenet

julia> A = Tensor(rand(2, 4, 2), (:i, :j, :k))
2×4×2 Tensor{Float64, 3, Array{Float64, 3}}:
[:, :, 1] =
 0.864098  0.504478  0.764864  0.576408
 0.17088   0.644568  0.889588  0.883436

[:, :, 2] =
 0.0782172  0.893183    0.507304  0.758659
 0.836397   0.00498142  0.103803  0.176029

julia> B = Tensor(rand(4, 2, 2), (:j, :i, :k))
4×2×2 Tensor{Float64, 3, Array{Float64, 3}}:
[:, :, 1] =
 0.932157  0.505743
 0.865242  0.1574
 0.290655  0.269668
 0.267595  0.609859

[:, :, 2] =
 0.675176  0.725486
 0.737603  0.377162
 0.254187  0.925577
 0.980923  0.207142

julia> A - B
ERROR: DimensionMismatch: dimensions must match: a has dims (Base.OneTo(2), Base.OneTo(4), Base.OneTo(2)), b has dims (Base.OneTo(4), Base.OneTo(2), Base.OneTo(2)), mismatch at 1
Stacktrace:
 [1] promote_shape
   @ ./indices.jl:178 [inlined]
 [2] promote_shape(a::Tensor{Float64, 3, Array{Float64, 3}}, b::Tensor{Float64, 3, Array{Float64, 3}})
   @ Base ./indices.jl:169
 [3] -(A::Tensor{Float64, 3, Array{Float64, 3}}, B::Tensor{Float64, 3, Array{Float64, 3}})
   @ Base ./arraymath.jl:7
 [4] top-level scope
   @ REPL[5]:1