DOC, CI: stats skewtest refguide failure

[tylerjereddy]

This is showing up in the CI of unrelated PRs like gh-14281 and appears to be related to small floating point precision difference (a bit outside my expertise, but pasting the raw log contents below):

021-06-22T22:42:06.0038137Z ===========
2021-06-22T22:42:06.0038731Z scipy.stats
2021-06-22T22:42:06.0039292Z ===========
2021-06-22T22:42:06.0039732Z 
2021-06-22T22:42:06.0040287Z scipy.stats.skewtest
2021-06-22T22:42:06.0044399Z --------------------
2021-06-22T22:42:06.0044659Z 
2021-06-22T22:42:06.0045354Z File "build/testenv/lib/python3.8/site-packages/scipy/stats/stats.py", line 1408, in skewtest
2021-06-22T22:42:06.0045792Z Failed example:
2021-06-22T22:42:06.0046136Z     skewtest([2, 8, 0, 4, 1, 9, 9, 0])
2021-06-22T22:42:06.0046476Z Expected:
2021-06-22T22:42:06.0046843Z     SkewtestResult(statistic=0.44626385374196975, pvalue=0.6554066631275459)
2021-06-22T22:42:06.0047232Z Got:
2021-06-22T22:42:06.0047587Z     SkewtestResult(statistic=0.4462638537419698, pvalue=0.6554066631275459)
2021-06-22T22:42:06.0047859Z 
2021-06-22T22:42:06.0047991Z 
2021-06-22T22:42:06.0048585Z File "build/testenv/lib/python3.8/site-packages/scipy/stats/stats.py", line 1410, in skewtest
2021-06-22T22:42:06.0049033Z Failed example:
2021-06-22T22:42:06.0049358Z     skewtest([1, 2, 3, 4, 5, 6, 7, 8000])
2021-06-22T22:42:06.0049693Z Expected:
2021-06-22T22:42:06.0050074Z     SkewtestResult(statistic=3.571773510360407, pvalue=0.0003545719905823133)
2021-06-22T22:42:06.0050439Z Got:
2021-06-22T22:42:06.0050811Z     SkewtestResult(statistic=3.5717735103604067, pvalue=0.00035457199058231423)
2021-06-22T22:42:06.0051078Z 
2021-06-22T22:42:06.0051208Z 
2021-06-22T22:42:06.0051811Z File "build/testenv/lib/python3.8/site-packages/scipy/stats/stats.py", line 1412, in skewtest
2021-06-22T22:42:06.0052251Z Failed example:
2021-06-22T22:42:06.0052613Z     skewtest([100, 100, 100, 100, 100, 100, 100, 101])
2021-06-22T22:42:06.0053231Z Expected:
2021-06-22T22:42:06.0053599Z     SkewtestResult(statistic=3.5717766638478072, pvalue=0.000354567720281634)
2021-06-22T22:42:06.0053977Z Got:
2021-06-22T22:42:06.0054335Z     SkewtestResult(statistic=3.5717766638478063, pvalue=0.00035456772028163574)
2021-06-22T22:42:06.0054615Z 
2021-06-22T22:42:06.0054745Z 
2021-06-22T22:42:06.0055035Z ERROR: refguide or doctests have errors

[ev-br]

There's an issue in refguide-check, I'll take a look.

[ev-br]

Locally a different one fails for me, but it's numerical noise anyway and hopefully gh-14283 makes it more robust by delegating these cases to np.allclose.