Nested arrays for simple slicing

[d7919]

This PR is mostly just for discussion -- it provides a sample implementation of a multi-dimensional container that supports simple slicing. It's ultimately backed by Array.

It's probably not very efficient currently.

It's missing unit tests.

Here's an ugly example of how it could be (ab)used

 ArrayND<BoutReal, 1> myArr(10);
  ArrayND<BoutReal, 2> myArr2(10, 20);
  ArrayND<BoutReal, 2> myArrPack(2, 5);

  // Fill a 1D array
  for(int i = 0 ; i<10; i++) {
    myArr[i] = i*i;
  }

 // Squash the 1D array into a 2D array.
  myArrPack.pack(myArr);

// Get a flattened (1D) version of a 2D array
  auto myArrTmp = myArrPack.flatten();

// Print all the values
  int count = 0;
  for (int i = 0 ; i<std::get<0>(myArrPack.shape()); i++) {
    for (int j = 0 ; j<std::get<1>(myArrPack.shape()); j++) {
      output<<myArr[count]<<" "<<myArrPack[i][j]<<" "<<myArrPack(i,j)<<" "<<myArrTmp(count)<<endl;
      count++;
    }
  }

// Get a slice of the 2D array -- this is a 1D array
auto mySlice = myArrPack[0];
ASSERT1(mySlice.size() == 5);
for (int i = 0; i<mySlice.size(); i++) {
   output<<i<<" (i+5)^2 = "<<mySlice[i]<<endl;
}

This could be useful for the inversion code where we deal with matrices/tensors and want to get the contiguous z-slices out of this data -- with ArrayND we can get these as Arrays without any copying.