[RFC]: add `ndarray/base/to-unflattened`

[kgryte]

This API is similar to https://docs.pytorch.org/docs/stable/generated/torch.unflatten.html, but instead of returning a view, we return a full copy to contiguous memory.

API signature:

function toUnflattened( x, dim, sizes )

The returned array should be a new array (e.g., use ndarray/base/empty-like to allocate before calling ndarray/base/assign). One will need to take care to properly compute the strides of the unflattened dimensions.

[kgryte]

This package is likely dependent on ndarray/base/unflatten as once you create an unflattened "base" view of x, it is a matter of using emptyLike followed by a call to assign/copy.