[NDTensors] More compact tensor printing

[mtfishman]

This enables more compact printing of ITensors, such as:

using ITensors: Index, random_itensor
d = 100
i, j = Index.((d, d))
a = random_itensor(i, j)

julia> show(a)
ITensor ord=2
Dim 1: (dim=100|id=386)
Dim 2: (dim=100|id=806)
NDTensors.Dense{Float64, Vector{Float64}}
 100×100
  1.1065511218675492    0.25558343779457593  …  -0.7563918255048953
 -0.19087995131196725   0.10936665763661539     -0.9511317856005679
 -0.1339119679780238   -0.9824888504596919      -0.8297798433753355
 -0.3380376950048613   -0.22502109940735232      1.238562637784077
  ⋮                                          ⋱  
  0.569878151816781     0.9722889212769971       0.15625262885769123
 -0.6617584401784992    1.1133126735067624       1.1441043736020045
 -1.2613730025434786   -0.3896145241325943       0.5647941511581129

which is achieved by setting limit = true when printing the tensor data as an array (see https://docs.julialang.org/en/v1/base/io-network/#Base.IOContext-Tuple{IO,%20Pair}).

Previously, all of the elements of the tensor were being printing, which filled the screen with output for larger tensors. This was an issue I came across in the ITensor CCQ school.

[mtfishman]

In the latest I also made some printing of ITensors from within arrays more compact by just printing the indices in those cases:

julia> [a a; a a]
2×2 Matrix{ITensor}:
 ((dim=100|id=386), (dim=100|id=806))  …  ((dim=100|id=386), (dim=100|id=806))
 ((dim=100|id=386), (dim=100|id=806))     ((dim=100|id=386), (dim=100|id=806))