BF(TST): parallel - take longer for producer to produce

[yarikoptic]

so that consumers have a better chance to finish up at least some
jobs (we got assert_greater_equal(len(results), 2) failing), before
the whole thing fails and we raise right away (that is what this
portion of the test tests).

If this does not show to be a reliable workaround, since we do not
use reraise_immediately=True mode yet anywhere, I gues we could just
disable this portion of the test.

Closes #5746

[codecov]

Codecov Report

Merging #5747 (83e2868) into maint (8c158c5) will decrease coverage by 7.56%.
The diff coverage is 87.50%.

❗ Current head 83e2868 differs from pull request most recent head e66e185. Consider uploading reports for the commit e66e185 to get more accurate results

@@            Coverage Diff             @@
##            maint    #5747      +/-   ##
==========================================
- Coverage   90.30%   82.74%   -7.57%     
==========================================
  Files         299      296       -3     
  Lines       42353    42334      -19     
==========================================
- Hits        38247    35029    -3218     
- Misses       4106     7305    +3199     

Impacted Files	Coverage Δ
datalad/core/distributed/clone.py	90.10% <ø> (-1.53%)	⬇️
datalad/core/distributed/push.py	88.52% <ø> (ø)
datalad/core/local/create.py	92.36% <ø> (-0.70%)	⬇️
datalad/core/local/diff.py	90.62% <ø> (ø)
datalad/core/local/run.py	98.23% <ø> (-0.89%)	⬇️
datalad/core/local/status.py	96.42% <ø> (ø)
datalad/distributed/create_sibling_ria.py	82.35% <ø> (ø)
datalad/distributed/export_archive_ora.py	81.53% <ø> (ø)
datalad/distribution/create_sibling.py	82.81% <ø> (-3.10%)	⬇️
datalad/distribution/create_sibling_github.py	84.37% <ø> (ø)
... and 174 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8c158c5...e66e185. Read the comment docs.

[yarikoptic]

eh, in case of the heavy logging we still get it to fail

======================================================================
FAIL: datalad.support.tests.test_parallel.test_gracefull_death
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/tmp/dl-miniconda-f10lrp_8/lib/python3.8/site-packages/nose/case.py", line 198, in runTest
    self.test(*self.arg)
  File "/tmp/dl-miniconda-f10lrp_8/lib/python3.8/site-packages/datalad/tests/utils.py", line 266, in _wrap_skip_if
    return func(*args, **kwargs)
  File "/tmp/dl-miniconda-f10lrp_8/lib/python3.8/site-packages/datalad/support/tests/test_parallel.py", line 202, in test_gracefull_death
    assert_greater_equal(8, len(results))
AssertionError: 8 not greater than or equal to 9

... with that effect -- I feel that I better just disable (I will comment out) that portion of the test so it stops annoying us ;-)

[yarikoptic]

to pacify Test for semver label / test (pull_request) where we test only for patch or internal and not tests (attn @jwodder), I added patch. Let's see now what would happen in the release workflow upon merge! BRACE FOR IMPACT!