Parallel notebook execution via multiprocessing spawn pool

[rnikutta]

Add -w/--workers CLI flag and workers notebook variable to run up to N notebooks simultaneously. Uses spawn context to avoid inheriting Jupyter's asyncio event loop into worker processes. Results stream in as each notebook finishes. Summary cell reports total CPU time, wall-clock time, and speedup — all derived from the returned testresults object. Default is workers=4.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[rnikutta]

@jacquesalice could you please review this PR? (not urgent). Run tests/testnotebooks.ipynb on gp13 with a few notebook directories specified in include. All should be the same as before, but these can now run with a number of parallel processes (workers=[number]). The printing is delayed until after each test is done (otherwise we'd have intermingled incomplete outputs). Thanks!

[jacquesalice]

@rnikutta this worked great! I tested it out by comparing the run speed of the following example with the old code vs this new code:

nbs = get_nbs(paths=paths,
              include=('/**/02_GettingStartedWithDataLab.ipynb',
                       '/**/ExploringSMASHDR2.ipynb',
                       '/**/DwarfGalaxiesInDelveDr2.ipynb',
                       '/**/GNIRS_DQS_DataAccessAtDataLab.ipynb',
                       '/**/LargeScaleStructureSdssCosmicSlime.ipynb',
                       '/**/ExtremelyMetalPoorStarsInTheLMC.ipynb'
                      ),
              exclude=exclude)

testresults = run(nbs, workers=workers)

The old code took 6min 19s to run, while the new code took only 3min 45s! PR Approved :-)

[rnikutta]

Great, many thanks @jacquesalice , I'll merge.