Use pycryptodome, allowing the use of pycrypto on GAE

[sirosen]

@mpdavis, is this method of GAE detection satisfactory?
I'm not an appengine user myself, so not ready to test, but I am interested in #43 getting merged.

Replaces / closes#43

[codecov-io]

Codecov Report

Merging #64 into master will decrease coverage by 0.11%.
The diff coverage is n/a.

@@ Coverage Diff @@## master #64 +/- ##
==========================================
- Coverage 94.53% 94.41% -0.12% 
==========================================
Files 12 12 Lines 841 841 ==========================================
- Hits 795 794 -1 - Misses 46 47 +1

Impacted Files	Coverage Δ
jose/backends/pycrypto_backend.py	96.15% <0%> (-0.97%)	⬇️

---

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 4e5cdbb...f1f46bf. Read the comment docs.

[sirosen]

I can mock over the google module and explicitly test this, to make codecov happier, but it seems that this integration is very shouty about unimportant changes in coverage.

@mpdavis, maybe you want to set a threshold in codecov.yaml? The doc for that file says it can be set to a %. Anything that decreases coverage by <1%, for example, should not be marked as failure, IMO.

[sirosen]

I've just validated that this installs correctly on GAE.
Also, I think this closes #41 ?

EDIT: I'm not actually sure whether or not my GAE validation was sufficient, being a novice in its use. I was able to install in the Google Cloud Shell, which I thought was the same environment as GAE, but I now notice that it actually installed pycryptodome, so maybe not...

[sirosen]

@mpdavis, a coworker just pointed out to me that pycrypto now fails to install on macOS 10.13
That elevates the importance of changing the default crypo lib off of pycrypto.

If this detection isn't satisfactory, I'm happy to swap it out for something else, but not being a GAE user it's hard to be confident in my testing.

[mpdavis]

That certainly would push us towards getting off of pycrypto as a default.

I can look into testing this and getting it in. We will need some way to handle this.

[mpdavis]

#57