Support ndarray deepcopy in C++ wrapper of AOT api

[ailzhang]

There're cases where users would want a deepcopy of an ndarray, so we can probably provide this as

auto arr_copy = arr.clone()

in our C++ wrapper. cc: @YuCrazing

[PENGUINLIONG]

What's the intent tho? Do you want an actual deep-copy or just something like np.zeros_like?

[YuCrazing]

In my case, I want an actual deep-copy. I would like to copy all data from ndarray A into ndarray B with the least codes. If we have a deep-copy interface, we do not need write codes like this:

auto B = allocate_ndarray(A.shape, A.elem_shape, ...);
memcpy_d2d(B, A);

[PENGUINLIONG]

hmm okay. there would be something like a clone.