Is there any interest in a nan-ignoring version of einsum a la nansum, nanprod, etc?  Any idea how difficult it would be to implement?

- Chris