[AG-35] Detect github repositories from github.com and github.io homepages

[felipead]

In order to improve license detection, it is imperative that we find a package's GitHub repository, since there's a very high change the license is going to be declared there.

We must parse github.com and github.io homepages, infering the repository url from these urls:

https://foo.github.com/bar => https://github.com/foo/bar
https://foo.github.io/bar => https://github.com/foo/bar

Here's an example. This Node.js package declares the homepage as:

  "homepage": "http://zaach.github.com/jsonlint/"

We can safely infer the associated GitHub repository, which is http://github.com/zaach/jsonlint. Inside this repository, the license can be inferred from parsing the README.md file, using the method described in #10.

This is very easy to do, as it is just a matter of adding another regular expression inside app/github/repository.py.