COMPAT/BLD: int dtype in json tests

[chris-b1]

Fixes failing windows build
https://ci.appveyor.com/project/jreback/pandas-465/build/job/p8hju341jyhnkogc

[jreback]

so maybe add a note in json docs that deserializstion will go to platform ints

[chris-b1]

The public json api (pd.read_json) actually coerces everything to int64, it's only the internal methods where you can get a platform int.

In [45]: s = pd.Series([1,2,3])

In [46]: pd.read_json(s.to_json(orient='records'), typ='series').dtype
Out[46]: dtype('int64')

In [47]: from pandas.json import loads

In [48]: loads(s.to_json(orient='records'), numpy=True).dtype
Out[48]: dtype('int32')

edit: one more example., just to confirm numpy=True also coerces

In [53]: pd.read_json(s.to_json(orient='records'), 
              typ='series', numpy=True, orient='records').dtype
Out[53]: dtype('int64')

[jreback]

ok thanks @chris-b1 ping on green.

[codecov-io]

Current coverage is 85.26% (diff: 100%)

Merging #14100 into master will not change coverage

@@             master     #14100   diff @@
==========================================
  Files           139        139          
  Lines         50502      50502          
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
  Hits          43063      43063          
  Misses         7439       7439          
  Partials          0          0          

Powered by Codecov. Last update a0151a7...a2825d9

[jreback]

thank you sir!