Less strict test, don't parallelise tests

[MarcoGorelli]

closes #4264

The test pymc3.tests.test_gp.TestMarginalKron passed during the CI process of #4262, but failed when that PR was merged and the CI process was run again (see here).

The test failure shows an absolute difference of 1e-6 - does this indicate a problem, or can the test just be made a bit less strict?

[codecov]

Codecov Report

Merging #4265 (8848620) into master (5ff9bbc) will increase coverage by 13.53%.
The diff coverage is n/a.

@@             Coverage Diff             @@
##           master    #4265       +/-   ##
===========================================
+ Coverage   73.90%   87.44%   +13.53%     
===========================================
  Files          88       88               
  Lines       14343    14343               
===========================================
+ Hits        10600    12542     +1942     
+ Misses       3743     1801     -1942     

Impacted Files	Coverage Δ
pymc3/model.py	88.88% <0.00%> (+0.22%)	⬆️
pymc3/backends/base.py	86.92% <0.00%> (+0.38%)	⬆️
pymc3/distributions/mixture.py	89.34% <0.00%> (+1.37%)	⬆️
pymc3/tests/helpers.py	60.34% <0.00%> (+1.72%)	⬆️
pymc3/backends/report.py	90.90% <0.00%> (+2.09%)	⬆️
pymc3/step_methods/hmc/base_hmc.py	91.66% <0.00%> (+2.50%)	⬆️
pymc3/tuning/scaling.py	58.18% <0.00%> (+5.45%)	⬆️
pymc3/step_methods/compound.py	95.65% <0.00%> (+8.69%)	⬆️
pymc3/distributions/distribution.py	94.31% <0.00%> (+8.78%)	⬆️
pymc3/distributions/continuous.py	93.27% <0.00%> (+9.15%)	⬆️
... and 26 more

[twiecki]

That's fine - thanks!

[twiecki]

Thanks @MarcoGorelli!