fix: improve version checking for nodes of workflows

[satra]

Reduce number of version checks for nodes of workflows run through multiprocessing .

This is a rather drastic step, and we should think through how to handle this in general. this will prevent this for other libraries which are used as well.

[effigies]

What if we add a new environment variable like NIPYPE_SKIP_ET?

[codecov]

Codecov Report

Merging #3152 into maint/1.4.x will increase coverage by 0.05%.
The diff coverage is 68.42%.

@@              Coverage Diff               @@
##           maint/1.4.x   #3152      +/-   ##
==============================================
+ Coverage        67.54%   67.6%   +0.05%     
==============================================
  Files              299     299              
  Lines            39499   39496       -3     
  Branches          5220    5220              
==============================================
+ Hits             26681   26701      +20     
+ Misses           12108   12087      -21     
+ Partials           710     708       -2

Flag	Coverage Δ
#smoketests	53.02% <45.45%> (-0.02%)	⬇️
#unittests	64.85% <55.26%> (+0.05%)	⬆️

Impacted Files	Coverage Δ
nipype/pipeline/plugins/tools.py	78.94% <ø> (ø)	⬆️
nipype/utils/__init__.py	100% <ø> (ø)	⬆️
nipype/utils/config.py	65% <ø> (-0.2%)	⬇️
nipype/algorithms/icc.py	57.53% <ø> (-0.58%)	⬇️
nipype/interfaces/dipy/preprocess.py	25.42% <0%> (-0.42%)	⬇️
nipype/interfaces/mrtrix/convert.py	17.18% <0%> (+0.17%)	⬆️
nipype/interfaces/dcmstack.py	35.1% <0%> (-0.27%)	⬇️
nipype/interfaces/fsl/epi.py	62.08% <0%> (-0.08%)	⬇️
nipype/algorithms/modelgen.py	67.6% <100%> (-0.07%)	⬇️
nipype/interfaces/spm/base.py	87% <100%> (ø)	⬆️
... and 10 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 39012c5...e2db192. Read the comment docs.

[effigies]

I think the logic is backwards in a couple places, and I'm a bit confused by the pyscript stuff.

[effigies]

This seems to work. See satra#31 for a test and a couple small suggestions.

Also, we currently support two other environment variables: NIPYPE_NO_MATLAB and NIPYPE_CONFIG_DIR. My inclination would be to stick with the NIPYPE_* convention. Unless you have a strong preference for NO_NIPYPE_ET?

[effigies]

Looking at the test failure, I'm unable to replicate it easily. It appears to be related to matplotlib/matplotlib#15832 and https://discourse.matplotlib.org/t/issues-with-my-installation-of-matplotlib/20746.

I can fiddle more later, but I wouldn't hold up merge for it.

Only remaining comment is that, as a bug fix, we should target maint/1.4.x and rebase on it.

Edit: Failing test fixed in maint/1.4.x. Rebasing should resolve it.

[effigies]

@satra Rebased cleanly, so I just force-pushed to keep this moving. Hope that's okay.