Respect byteorder in StataReader

[aplanas]

Partially fix #11282 for s390x

[jreback]

would love to include, but can you include a test. ideally write a test frame in (both little/big) endian then read back.

[aplanas]

Actually I detected that because of this test:

nosetests -s pandas/io/tests/test_stata.py:TestStata.test_read_dta18

This test was OK in x86_64 and wrong on ARM and S390X machines, now is green in both.

[jreback]

don't delete tests
better to skip that on platforms where it is inappropriate (and use a different one where it is)

so ideally have a file in a different byte order then test it can be read in the correct platform (and errors on incorrect platforms)

[kshedden]

Does anyone have access to Stata on a BE machine? If so we should just convert a few of the test files to BE and then test that we can read them back in on a LE machine.

I don't have access to a BE machine with Stata, but I can provide Stata code to load/save the files. If run on a BE machine that should convert the file byteorder to BE. I don't think Stata will save files in non-native byte order, so I can't do this on any of my machines.

[jreback]

closing, but pls reopen if you can add tests