[BUG] Array proxy in cudf.pandas don't include special casing for ndarray.flat

[shwina]

The arr.flat attribute should return a flatiter object, but they currently return a generator. Unlike flatiter, generator objects cannot be written to:

In [1]: %load_ext cudf.pandas

In [2]: import pandas as pd

In [3]: s = pd.Series([1,2 , 3])

In [4]: arr = s.values

In [5]: arr.flat
Out[5]: <generator object _maybe_wrap_result.<locals>.<genexpr> at 0x7f70a82cc310>

In contrast:

import numpy as np

arr = np.arange(5)
print(type(arr.flat))
arr.flat[:3] = 100
print(arr)
<class 'numpy.flatiter'>
[100 100 100   3   4]

import cupy as cp
arr = cp.arange(5)
print(type(arr.flat))
arr.flat[:3] = 100
print(arr)
<class 'cupy._indexing.iterate.flatiter'>
[100 100 100   3   4]