Implement DataFrame.map_in_pandas

[HyukjinKwon]

This PR implements a Koalas specific API in order to directly use pandas DataFrame APIs.

For pandas Series APIs, we have a workaround for instance:

>>> ks.range(10)[['id']].apply(lambda series: series)['id']
0    0
1    1
2    2
3    3
4    4
5    5
6    6
7    7
8    8
9    9
Name: id, dtype: int64

however, there is no way to use the APIs in the pandas DataFrame.

There is a similar API called map_partitions in Dask and this API is inspired by it.

[codecov-io]

Codecov Report

Merging #1276 into master will decrease coverage by 0.06%.
The diff coverage is 92.85%.

@@            Coverage Diff             @@
##           master    #1276      +/-   ##
==========================================
- Coverage   95.15%   95.09%   -0.07%     
==========================================
  Files          34       34              
  Lines        7208     7236      +28     
==========================================
+ Hits         6859     6881      +22     
- Misses        349      355       +6

Impacted Files	Coverage Δ
databricks/koalas/frame.py	96.47% <92.85%> (-0.06%)	⬇️
databricks/koalas/usage_logging/__init__.py	97.29% <0%> (ø)	⬆️
databricks/koalas/usage_logging/usage_logger.py	100% <0%> (ø)	⬆️
databricks/koalas/__init__.py	85.1% <0%> (-8.52%)	⬇️
databricks/conftest.py	96.22% <0%> (ø)	⬆️
databricks/koalas/namespace.py	87.87% <0%> (ø)	⬆️
databricks/koalas/plot.py	94.28% <0%> (ø)	⬆️
databricks/koalas/testing/utils.py	78.51% <0%> (ø)	⬆️
databricks/koalas/groupby.py	91.43% <0%> (ø)	⬆️
... and 1 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6662ae6...fa8db4d. Read the comment docs.

[HyukjinKwon]

@ueshin, let me merge this for now. The implementation is similar with DataFrame.apply so there won't be too many things to closely review.

Also, this PR has a reference from Dask, it allows many workarounds for example:

        >>> def take_func(pdf) -> ks.DataFrame[int, int]:
        ...     return pdf.query('A == 1')
        >>> df.map_in_pandas(take_func)

Let me know if you have any concern!