__iter__() for Index & MultiIndex

[itholic]

This PR proposes __iter__() for Index & MultiIndex by using toLocalIterator like we did at DataFrame.iterrows and Series.iteritems.

>>> kidx = ks.Index([1, 2, 3])
>>> for value in kidx:
...     value
...
1
2
3

[HyukjinKwon]

I think these were explicitly unsupported per #836

[itholic]

@HyukjinKwon , Thanks for the comment.

I just read the issue, but maybe I think the topic of that issue is more closer to "Koalas should returning ks.Series rather than np.array for pandas APIs that returning np.array" rather than "Koalas shouldn't support the __iter__ for ks.Series/ks.Index" ?

Because I think it is not that danger to just iterate the items in ks.Series or ks.Index one by one, unlike returning the whole item set in a ks.Series or ks.Index at once like pd.Series.unique().

What do you think about ?

[ueshin]

If we support it, users can unconsciously do list(kser) or [ do something for row in kser ] which cause OOM so easily.

[itholic]

@ueshin Thanks! I'd close this now :)