Implemented DataFrame.lookup

[itholic]

This PR proposes DataFrame.lookup.

>>> kdf = ks.DataFrame({'A': [3, 4, 5, 6, 7],
...                     'B': [10.0, 20.0, 30.0, 40.0, 50.0],
...                     'C': ['a', 'b', 'c', 'd', 'e']})
>>> kdf
   A     B  C
0  3  10.0  a
1  4  20.0  b
2  5  30.0  c
3  6  40.0  d
4  7  50.0  e

>>> kdf.lookup([0], ["C"])
array(['a'], dtype=object)

>>> kdf.lookup([2, 3], ["A", "B"])
array([ 5., 40.])

[itholic]

Thanks, @HyukjinKwon

[itholic]

Thanks, @ueshin

[HyukjinKwon]

Does this mean we launch jobs as many as labels?

[xinrong-meng]

Out of curiosity, why did we change it to np.asarray(pd.Series(lookups))?

[xinrong-meng]

https://issues.apache.org/jira/browse/SPARK-36434