Handle the case when arraystruct->meminfo is null to close gh-965

[oleksandr-pavlyk]

Closes gh-965 by raising Python exception when arraystruct->meminfo field is null, and unhandled exception would have caused dereferencing of the null pointer and subsequent crash.

• Have you provided a meaningful PR description?
• Have you added a test, reproducer or referred to an issue with a reproducer?
• Have you tested your changes locally for CPU and GPU devices?
• Have you made sure that new changes do not introduce compiler warnings?
• If this PR is a work in progress, are you filing the PR as a draft?

[chudur-budur]

LGTM

[diptorupd]

@oleksandr-pavlyk @chudur-budur the solution is fine for now especially as it has already been merged.

I wanted to implement it in a different way. Look at what Numba does for NumPy arrays: https://github.com/numba/numba/blob/ce69f3010c77942123815d86900cb841c0e5d2fa/numba/core/runtime/_nrt_python.c#L366

Numba uses the arraystruct.meminfo if the meminfo object is populated. Otherwise it lets the ndarray base to be NULL when calling PyArray_NewFromDescr.

    array = (PyArrayObject *) PyArray_NewFromDescr(retty, descr, ndim,
                                                   shape, strides, arystruct->data,
                                                   flags, (PyObject *) miobj);

We should have equivalent C-API for tensor and dpnp and handle the case where we do not have meminfo.

[oleksandr-pavlyk]

@diptorupd I think it should be reimplemented as needed. The fix went in like this to comply with SDL requirements.