CI: Mark slow tests as slow to speed up ci

[phofl]

We have a few tests which are really slow and not marked as such. I went through the list below and marked everything greater than 4 seconds as slow. Should give us between 4 and 5 minutes performance boost

# 29.68s call pandas/tests/io/sas/test_sas7bdat.py::TestSAS7BDAT::test_iterator_loop # 12.98s call pandas/tests/test_sorting.py::TestSorting::test_int64_overflow_moar # 11.43s call pandas/tests/indexing/multiindex/test_chaining_and_caching.py::test_indexer_caching # 10.13s call pandas/tests/groupby/transform/test_transform.py::test_cython_transform_frame[shift-args3-] # 9.87s call pandas/tests/io/sas/test_sas7bdat.py::TestSAS7BDAT::test_from_iterator # 9.79s call pandas/tests/groupby/transform/test_transform.py::test_cython_transform_frame[shift-args2-] # 9.41s teardown pandas/tests/window/moments/test_moments_rolling_quantile.py::test_center_reindex_frame[1.0] # 8.82s call pandas/tests/io/sas/test_xport.py::TestXport::test1_basic # 8.43s call pandas/tests/io/parser/common/test_chunksize.py::test_warn_if_chunks_have_mismatched_type[python] # 8.38s call pandas/tests/io/parser/common/test_chunksize.py::test_chunks_have_consistent_numerical_type[python] # 8.13s call pandas/tests/io/parser/test_c_parser_only.py::test_grow_boundary_at_cap[c_low] # 8.12s call pandas/tests/io/parser/test_c_parser_only.py::test_grow_boundary_at_cap[c_high] # 6.97s call pandas/tests/indexes/datetimes/test_date_range.py::TestDateRanges::test_date_range_int64_overflow_stride_endpoint_different_signs # 5.90s call pandas/tests/frame/test_constructors.py::TestDataFrameConstructors::test_constructor_mrecarray # 5.84s teardown pandas/tests/window/moments/test_moments_rolling_skew_kurt.py::test_center_reindex_frame[skew] # 5.61s call pandas/tests/io/parser/test_c_parser_only.py::test_precise_conversion[c_high] # 5.29s call pandas/tests/io/sas/test_sas7bdat.py::TestSAS7BDAT::test_path_pathlib # 5.13s call pandas/tests/groupby/transform/test_transform.py::test_cython_transform_frame[cumsum-args1-] # 4.93s call pandas/tests/io/sas/test_sas7bdat.py::TestSAS7BDAT::test_from_file # 4.86s call pandas/tests/io/sas/test_sas7bdat.py::TestSAS7BDAT::test_from_buffer # 4.84s call pandas/tests/io/sas/test_sas7bdat.py::TestSAS7BDAT::test_path_localpath # 4.59s call pandas/tests/indexes/multi/test_integrity.py::test_hash_collisions # 4.26s call pandas/tests/indexing/interval/test_interval.py::TestIntervalIndex::test_loc_getitem_large_series # 3.99s call pandas/tests/io/parser/test_c_parser_only.py::test_precise_conversion[c_low]

[jbrockmendel]

some of these occasionally fail for unknown reasons, e.g. im pretty sure ive seen test_warn_if_chunks_have_mismatched_type and test_int64_overflow_moar in the logs. can we keep those? or possibly (i know this is much more labor-intensive) try to trim down these tests?

[phofl]

Do you know if test_warn_if_chunks_have_mismatched_type fails for the python or c engine? The python is the slow one

Added them back in

[jbrockmendel]

Do you know if test_warn_if_chunks_have_mismatched_type fails for the python or c engine? The python is the slow one

not off the top of my head, no

[jreback]

thanks @phofl