Fix tail to use the resolved copy

[itholic]

When use tail(), the result is not correct after modifying the data.

>>> kdf = ks.DataFrame({"A": [1, 2, 3]})
>>> kdf.tail(2)
   A
1  2
2  3
>>> (kdf + 1).tail(2)
   A
1  2
2  3
>>> (kdf.to_pandas() + 1).tail(2)
   A
1  3
2  4

This should resolve #1941

[itholic]

I'm not sure maybe should we do this only for Series.tail() when the given column name is not exists in given spark_frame ?

[codecov-io]

Codecov Report

Merging #1942 (fda74df) into master (809ed86) will decrease coverage by 0.00%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #1942      +/-   ##
==========================================
- Coverage   94.16%   94.15%   -0.01%     
==========================================
  Files          41       41              
  Lines       10055    10055              
==========================================
- Hits         9468     9467       -1     
- Misses        587      588       +1     

Impacted Files	Coverage Δ
databricks/koalas/series.py	97.04% <100.00%> (ø)
databricks/koalas/namespace.py	84.02% <0.00%> (-0.21%)	⬇️

---

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 809ed86...478ec20. Read the comment docs.

[ueshin]

The behavior described in the description is totally expected.

I guess the problem is in DataFrame.tail because it happens even there.

>>> kdf = ks.DataFrame({"A": [1, 2, 3]})
>>> kdf.tail(2)
   A
1  2
2  3
>>> (kdf + 1).tail(2)
   A
1  2
2  3
>>> (kdf.to_pandas() + 1).tail(2)
   A
1  3
2  4

[itholic]

@ueshin Thanks! Let me add a related tests and take a look at this again