Use F.datediff() for subtraction of dates as a workaround.

[ueshin]

Since currently we don't support timedelta type in pandas, we haven't support subtraction of dates, but we already have a workaround for subtraction of timestamps returning long instead of timedelta. We can also support subtraction of dates returning int.
We should revisit after we support timedelta type.

[codecov-io]

Codecov Report

Merging #1439 into master will increase coverage by 0.07%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #1439      +/-   ##
==========================================
+ Coverage   94.83%   94.91%   +0.07%     
==========================================
  Files          34       34              
  Lines        8078     8082       +4     
==========================================
+ Hits         7661     7671      +10     
+ Misses        417      411       -6     

Impacted Files	Coverage Δ
databricks/koalas/base.py	98.06% <100.00%> (+0.42%)	⬆️
databricks/koalas/frame.py	96.43% <0.00%> (+0.20%)	⬆️
databricks/koalas/datetimes.py	99.11% <0.00%> (+0.88%)	⬆️

---

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 9d97ffc...073156b. Read the comment docs.