Implement pure python rsa signing based on rsa module

[zejn]

This offers a fallback and ensures the library is usable even when a supported fast C implementation isn't available.

[codecov-io]

Codecov Report

Merging #58 into master will decrease coverage by 0.07%.
The diff coverage is 93.67%.

@@            Coverage Diff             @@
##           master      #58      +/-   ##
==========================================
- Coverage   93.69%   93.62%   -0.08%     
==========================================
  Files          12       13       +1     
  Lines         841     1004     +163     
==========================================
+ Hits          788      940     +152     
- Misses         53       64      +11

Impacted Files	Coverage Δ
jose/backends/__init__.py	36.36% <0%> (-13.64%)	⬇️
jose/backends/cryptography_backend.py	97.92% <84.61%> (-2.08%)	⬇️
jose/backends/pycrypto_backend.py	96.36% <92.85%> (+0.2%)	⬆️
jose/backends/rsa_backend.py	97.2% <97.2%> (ø)

---

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 28cc671...873807d. Read the comment docs.

[mpdavis]

I need to go over this with a finer comb, but I wanted to let you know that I agree with the general idea here.

[mpdavis]

My apologies, but #52 caused some conflicts with the tests due to a change to parameterize those tests, which should avoid some copy-paste for you.

Additionally, it added the ability to export keys to JWKs which will need to be implemented on the new RSA key.

[zejn]

Sure, will try to give some more attention to this in a timely manner.

[zejn]

I've updated this pull request to finally implement RSA using pure Python package "rsa".
The RSAKey.to_pem methods are updated to allow exporting keys in PKCS#1 and PKCS#8 format for private keys and PKCS#1 and SubjectPublicKeyInfo for public keys.
This pull request also changes the default from package "pycryptodome" to "rsa" and thus avoids having issues with namespace conflicts between pycrypto and pycryptodome.

[mpdavis]

I finally took a look at this, and everything looks good to me.