Ci #34131 rm slow fixtures

[OlivierLuG]

• [ x ] Part of CI/TST: slow tests #34131
• [ x ] passes black pandas
• [ x ] passes git diff upstream/master -u -- "*.py" | flake8 --diff
• [ FAIL ] tests passed for 11 of 12 files changed. 6 tests fail inside the file pandas/core/arrays/datetimes.py , but there are not concerned by this PR. So I decide to push anyway.

I went through tests tagged with @pytest.mark.slow and take a look at the duration. The test is unmarked if the test duration is below 1 second.
note: I considered the total time for each combination in case of parameters

A total of 76 flags slow were unmark, so we can focus on "real" slow tests.

[jbrockmendel]

for future reference, lets refer to these as pytest.mark.slow, since they're not technically fixtures

[jbrockmendel]

looks like some of the failing tests need a td.skip_if_no_scipy decorator

[jreback]

@OlivierLuG aplaud the effort here. this is a bit tricky because IIRC we only run the slow on a single job (on travis); the issue is that some of these tests are marked slow not to indicate they are slow but rather that we want to run them there, e.g. plotting tests and such.

we have a pretty tricky setup with this, so you will have to play with the ci to get this to pass.

you might want to start with a smaller scope (e.g. start with x tests), then get it passing, then keep adding.

[OlivierLuG]

Thanks for the explaination. If I do understand correctly, the objective of this issue is either to unmark the slow tests, and still passing the CI tests or to optimise the speed of the tests itself ?

[jreback]

i think a good start here would be to fix some of the really slow tests themselves

also the original issue was about some fast tests that are actually slow on ARM