No-copy version of PyReverseDims for NumPy to Julia arrays

It would be great if there was an equivalent of `PyReverseDims` for the other direction, e.g. a method that takes a `PyArray` and returns a regular Julia Array with the dimensions inverted. For things like images from PIL, the numpy array of size `(nvertical, nhorizontal, ncolor)` mostly comes from NumPy arrays being row-major. In Julia, an array of size `(ncolor, nhorizontal, nvertical)` also feels natural because Julia is column-major. These two arrays would have the same representation in memory, but I cannot figure out how to accomplish this conversion without copying the underlying data, even when using `reinterpret` or `resize`.

Start by reviewing the existing PyReverseDims behavior and the PyArray conversion APIs described in the issue. Determine whether a regular Julia Array can expose the inverted dimensions without copying the underlying data; done means providing the requested no-copy conversion for NumPy arrays, including the image-shaped example.

Scritto dal modello di indicizzazione a partire dal testo della issue.