Implement additional methods for CPD

[dahong67]

Some options:

• destructuring, e.g., λ, (U1, U2, U3) = M where M::CPD
  For reference, this is how it's done for SVD:

  Base.iterate(S::SVD) = (S.U, Val(:S))
  Base.iterate(S::SVD, ::Val{:S}) = (S.S, Val(:V))
  Base.iterate(S::SVD, ::Val{:V}) = (S.V, Val(:done))
  Base.iterate(S::SVD, ::Val{:done}) = nothing

• norm(M::CPD) = sqrt(sum(abs2,M[I] for I in CartesianIndices(size(M))))
  • Add norm method for CPD #44
• *(a::Number, M::CPD)
• /(M::CPD, a::Number)
• normalize(M::CPD) = M/norm(M)
• iterator over components (maybe spelled eachcomp) analogous to eachrow, eachcol, eachslice.
• permuting/sorting the components (analogous to arrange in Tensor Toolbox)
  • Add permutecomps/permutecomps! #66
• permutedims (analogous to permute in Tensor Toolbox)
• getcomp (analogous to getindex for AbstractArrays. see also extract in Tensor Toolbox)
• Normalize components #64
• Conversion to Array #57

Implementations in other packages/toolboxes:

• Tensor Toolbox: https://gitlab.com/tensors/tensor_toolbox/-/tree/dev/@ktensor
• Tensorly:
  • http://tensorly.org/stable/modules/api.html#module-tensorly.cp_tensor
  • http://tensorly.org/stable/_modules/tensorly/cp_tensor.html