`ak.array` gives unexpected results on a transposed numpy multi-dimensional array.

[ajpotts]

Describe the bug
ak.array gives unexpected results on a transposed numpy multi-dimensional array.

To Reproduce

In [81]: 
    ...: import numpy as np
    ...: rows = 5
    ...: cols = 5
    ...: nda = np.random.randint(1, 10, (rows, cols))
    ...: nda
Out[81]: 
array([[7, 7, 7, 8, 6],
       [9, 4, 9, 6, 2],
       [2, 2, 3, 1, 9],
       [2, 7, 1, 3, 7],
       [4, 2, 7, 5, 1]])

In [82]: 
    ...: nda_to_ak = ak.array(nda)
    ...: nda_to_ak
    ...: 
Out[82]: array([array([7 7 7 8 6]) array([9 4 9 6 2]) array([2 2 3 1 9]) array([2 7 1 3 7]) array([4 2 7 5 1])])

In [83]: 
    ...: npa = np.transpose(nda)
    ...: npa
Out[83]: 
array([[7, 9, 2, 2, 4],
       [7, 4, 2, 7, 2],
       [7, 9, 3, 1, 7],
       [8, 6, 1, 3, 5],
       [6, 2, 9, 7, 1]])

In [84]: 
    ...: npa_to_ak = ak.array(npa)
    ...: npa_to_ak
    ...: 
    ...: 
Out[84]: array([array([7 7 7 8 6]) array([9 4 9 6 2]) array([2 2 3 1 9]) array([2 7 1 3 7]) array([4 2 7 5 1])])

[ajpotts]

This is interesting:

In [32]: ak.array(nda_transpose)
Out[32]: array([array([1 1 3 6 7]) array([4 6 6 8 1]) array([3 7 2 4 8]) array([5 6 5 9 1]) array([3 2 7 9 6])])

In [33]: ak.array(nda_transpose.copy())
Out[33]: array([array([1 4 3 5 3]) array([1 6 7 6 2]) array([3 6 2 5 7]) array([6 8 4 9 9]) array([7 1 8 1 6])])

In [34]: nda_transpose
Out[34]: 
array([[1, 4, 3, 5, 3],
       [1, 6, 7, 6, 2],
       [3, 6, 2, 5, 7],
       [6, 8, 4, 9, 9],
       [7, 1, 8, 1, 6]])

[stress-tess]

Dropped a comment on the PR. The reason for this is that for numpy the transpose is being treated as a different view of the same data (switches between column and row major), but the underlying data hasn't changed. We send the underlying data over as a buffer and ignore the order. When we do a copy, it actually moves the data elsewhere

This comment of the PR explains better and looks at the flags of the array and transpose:
#3761 (comment)