Add `NDIterator<T>` overload with support for specific axis.

[Nucs]

NDIterator is useful, we should add an overload that handles specific axis iterator:
usage:

new NDIterator<T>(ndarray, axis: 1);

Also add an overload to extension ndarray.AsIterator<T>(axis: 1);

[Nucs]

Related NEPs: See #548 for [NEP10 NEP38 NEP54] - NEP 10 covers iterator optimization including axis-specific iteration patterns.

[Nucs]

#611 Closes this

[Nucs]

Following up on my earlier note (#611 gave us the engine): 0.70.0 (#628) adds np.nested_iters for per-axis-subset nested loops, plus the order-aware typed np.nditer<T>:

var a = np.arange(6).reshape(2, 3);

// one iterator per axis group; advancing the outer re-bases the inner
var iters = np.nested_iters(a, new[] { new[] { 0 }, new[] { 1 } });
Console.WriteLine(iters.Length);   // 2

Closing.