Incorrect construction of SimpleArray from a discontiguous slice of ndarray

[xingularity]

When a SimpleArray is constructed through the "array" argument in Python and the numpy array passed to this argument is a subset using ndarray accessor, the SimpleArray is constructed with wrong section of data from the original ndarray. Following code is a demonstration.

from modmesh import SimpleArrayFloat64
import numpy as np
a = np.array([[1,2,3,4,5], [10,20,30,40,50],[100,200,300,400,500]])

Now we have

array([[  1,   2,   3,   4,   5],
       [ 10,  20,  30,  40,  50],
       [100, 200, 300, 400, 500]])

If we construct a SimpleArray with

b = SimpleArrayFloat64(array=a[:, 4])
print(b.ndarray)

The print result is [ 5., 10., 20.].
We should expect a SimpleArray contains [5, 50, 500] instead of [ 5., 10., 20.]

[yungyuc]

The SimpleArray constructor expects a contiguous memory buffer and disregard the strides set in ndarray. Before SimpleArray supports the same stride as ndarray, it should throw an exception when the input ndarray is not contiguous.

For now, please use b = SimpleArrayFloat64(array=a[:,4].copy()) to work around.

[xingularity]

Thanks @yungyuc, I already used that workaround in my code.

[yungyuc]

@tigercosmos Would you have time to take a look?

[tigercosmos]

I can help support this starting from the beginning of November.

[yungyuc]

Fixed by #438