Fallback to self = np.ndarray when necessary

[bryevdv]

This PR renames the clone_class decorator to clone_np_ndarray to reflect the fact that it is closely tied to the implementation of np.ndarray.

Additionally, this PR fixes the issue reported by @fduguet-nv

>>> import cunumeric as num
>>> ones = num.ones((10,))
>>> ones.mean()
array(1.)
>>> ones.std()
<console>:1: RuntimeWarning: cuNumeric has not implemented numpy.ndarray.std and is falling back to canonical numpy. You may notice significantly decreased performance for this function call.
Traceback (most recent call last):
  File "<console>", line 1, in <module>
  File "/home/fduguet/github/cunumeric-fduguet/cunumeric/coverage.py", line 165, in wrapper
    return func(*args, **kwargs)
TypeError: descriptor 'std' for 'numpy.ndarray' objects doesn't apply to a 'ndarray' object

A fix similar to #430 was added, which is tested in a new integration test. The fix is slightly more general, which would allow for any future specific clone_xyz class decorators to also make use of it. This PR supersedes #430.

[bryevdv]

@magnatelee I don't love this but I am not sure what else we can really do.