Use cPickle on Python 2 if available

[micbou]

We should use the cPickle module if available since it is a C implementation of pickle that is way faster.

Here are the difference between pickle and cPickle measured with this script when completing the os, numpy, and cv2 (opencv-python) modules on Ubuntu 18.04 64-bit:

Interpreter	Python 2.7				Python 3.6
Env.	Python 2.7		Python 3.6		Python 2.7		Python 3.6
Module	pickle	cPickle	pickle	cPickle	pickle	cPickle	pickle	cPickle
os	0.17s	0.107s	0.18s	0.156s	0.0946s	0.0741s	0.0738s	0.0728s
numpy	0.245s	0.199s	0.261s	0.242s	0.209s	0.184s	0.181s	0.179s
cv2	0.924s	0.652s	1.05s	0.932s	0.519s	0.355s	0.471s	0.47s

We can see a significant improvement for all combinations of Python except when the interpreter running Jedi and the environment are both Python 3 (which is expected since there is no cPickle on Python 3).

---

This change is 

[codecov-io]

Codecov Report

Merging #1151 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #1151   +/-   ##
=======================================
  Coverage   92.78%   92.78%           
=======================================
  Files          57       57           
  Lines        6902     6902           
=======================================
  Hits         6404     6404           
  Misses        498      498

---

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 e96ebbe...fabd8d5. Read the comment docs.

[davidhalter]

I'm a bit surprised that Python3 is faster with importing cPickle. cPickle doesn't even exist there. This pull request should therefore not change anything.

[davidhalter]

Oh, I realize now that it makes sense that it's faster if you run the interpreter as Python2. Sorry my bad.

Thanks for researching this!