Prevent spurious cache clearing when calling a cached func inside a Parallel call

[pierreglaser]

Fixes #1035

cc @ogrisel - I need to test a few more cases, but it should be reviewable.

[codecov]

Codecov Report

Merging #1093 into master will decrease coverage by 0.61%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #1093      +/-   ##
==========================================
- Coverage   94.52%   93.91%   -0.62%     
==========================================
  Files          47       47              
  Lines        6910     6955      +45     
==========================================
  Hits         6532     6532              
- Misses        378      423      +45     

Impacted Files	Coverage Δ
joblib/func_inspect.py	91.01% <ø> (-0.60%)	⬇️
joblib/memory.py	95.56% <100.00%> (+0.16%)	⬆️
joblib/test/test_memory.py	98.41% <100.00%> (-0.08%)	⬇️
joblib/backports.py	44.73% <0.00%> (-39.48%)	⬇️
joblib/test/test_store_backends.py	91.42% <0.00%> (-5.72%)	⬇️
joblib/_memmapping_reducer.py	94.33% <0.00%> (-2.27%)	⬇️
joblib/test/test_memmapping.py	97.33% <0.00%> (-1.91%)	⬇️
joblib/pool.py	86.17% <0.00%> (-1.63%)	⬇️
joblib/disk.py	90.47% <0.00%> (-1.59%)	⬇️
joblib/_parallel_backends.py	94.92% <0.00%> (-1.57%)	⬇️
... and 4 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 bc8c8cf...4438013. Read the comment docs.

[pierreglaser]

@ogrisel feel free to take a look, this is reviewable :)

[ogrisel]

LGTM!

[ogrisel]

That's a good question. It does not seem to be required to fix #1035 though (I tried manually with the reproducer in an interactive ipython session).

[ogrisel]

I also tried to stop and restart the ipython session and the cache survived.

[ogrisel]

Thank you very much @pierreglaser . Actually, I forgot to update the changelog prior to merging this. I will do it now.