Modern multiprocessing for MCMC NEGFC, small fix to tutorial 05A

[IainHammond]

So emcee has been locked into version 2.2.1 for some time, which was release on 15 July 2016. The reason was likely to be due to the change in multiprocessing behaviour in version 3.0 -> with the removal of the threads argument in favour of pool in multiprocessing.

From https://emcee.readthedocs.io/en/stable/user/upgrade/
"The threads keyword argument has been removed in favor of the pool interface (see the Parallelization tutorial for more information). The old interface had issues with memory consumption and hanging processes when the sampler object wasn’t explicitly deleted. The pool interface has been supported since the first release of emcee and existing code should be easy to update following the Parallelization tutorial."

This older version was indeed causing some occasional problems (at least for me) such as memory accumulating or the code seemingly stopping. To overcome this I've jumped us up to emcee 3.1.6 and made all of the recommended changes to the call for emcee.EnsembleSampler provided in the docs. It works, and it produces the same results. Is it faster? A bit. The user doesn't need to manually assign the number of processors anymore either as python automatically determines what is available. The multiprocessing start method is forkserver by default as it was running about 10% faster than fork. If it's unavailable, spawn is used instead so that Windows is supported (albeit spawn is about 20% slower than forkserver). Finally, a new moves keyword replaces the a keyword, and sampler.lnprobability is now sampler.get_log_prob().

The implication of this is that VIP technically needs emcee 3.0 -> from now on and requirements.txt has been updated to remove the version lock. In the long term this is much better, but will cause an error if a user doesnt update emcee following this pull request.

Random things:

• fixed a wrong keyword argument in tutorial 05A
• fixed a deprecated ipython import

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

Attention: Patch coverage is 1.08696% with 91 lines in your changes missing coverage. Please review.

Project coverage is 23.77%. Comparing base (4a8e9cf) to head (8d0ab52).
Report is 5 commits behind head on master.

Files with missing lines	Patch %	Lines
vip_hci/fm/negfc_mcmc.py	1.09%	91 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #658      +/-   ##
==========================================
- Coverage   33.13%   23.77%   -9.35%     
==========================================
  Files          87       87              
  Lines       16696    16720      +24     
==========================================
- Hits         5531     3975    -1556     
- Misses      11165    12745    +1580     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[VChristiaens]

Thanks for the changes @IainHammond. I think it'd still better for the user to be able to provide the number of CPUs to use for the MCMC. In many scenarios, either the user has to share the machine with others, or may want to run other jobs on their machine simultaneously, so would not want all available CPUs to be used. Can you adapt this?

[IainHammond]

good call - provided nproc to Pool in the latest commit (8d0ab52)

[VChristiaens]

Thanks @IainHammond !