repr and repr_html improvements

[rxin]

1. Fixed the bug that DataFrame. __repr__ was turning the entire DataFrame into pandas DataFrame, rather than simply taking the top 1000 rows.

2. Added a cache so repeated calls of repr and repr_html don't trigger multiple Spark actions. But then I undid that because cache invalidation is difficult. I did leave some of the infrastructure still in place and added test cases to make sure repr and repr_html are correct when it comes to cache invalidation, if we ever implement any form of caching.

[rxin]

Eh the test cases fail due to an existing flakiness issue: sort_values are not deterministic when there are two elements that are the same.

[codecov-io]

Codecov Report

Merging #258 into master will increase coverage by 0.16%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #258      +/-   ##
==========================================
+ Coverage   92.19%   92.35%   +0.16%     
==========================================
  Files          35       35              
  Lines        3164     3205      +41     
==========================================
+ Hits         2917     2960      +43     
+ Misses        247      245       -2

Impacted Files	Coverage Δ
databricks/koalas/generic.py	95% <ø> (ø)	⬆️
databricks/koalas/series.py	91.28% <100%> (ø)	⬆️
databricks/koalas/tests/test_utils.py	100% <100%> (ø)	⬆️
databricks/koalas/utils.py	100% <100%> (ø)	⬆️
databricks/koalas/tests/test_dataframe.py	100% <100%> (ø)	⬆️
databricks/koalas/tests/test_series.py	100% <100%> (ø)	⬆️
databricks/koalas/frame.py	92.43% <100%> (+0.48%)	⬆️

---

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 60b123c...6912754. Read the comment docs.