Emails are now fetched from diaspora and added to
synapse automatically \o/

Also, new release time!

mail@shamilk.me has been lost. So moving to a more
stable email address.

We are already running `synapse-diaspora-auth` in production
on diasp.in and poddery.com so moving this to a Production classifier
would be good.

Check password fails when a non-ascii character is in the password.
Fixed that

Some error with pymysql (perhaps?) causes poddery.com
to only accept 1 or 2 auth requests before failing.
This should fix that. (probably)

Before, v.0.0.3, the connection was reinitialized everytime a new
auth request was sent to the module. But after that version, things
changed and a single connection was used.

This caused #2. It was because the transactions weren't commited
before querying again. Which stops the auth provider from getting
new data. So, the context manager is also used on the connection
everytime a query is sent to ensure the transactions are committed.

So, hopefully this commit fixes #2

A new engine variable is added to support switching of the
database engine. Currently supports postgresql and mysql databases.

This should be enough as mysql and postgres are the only 2 database
engines supported by diaspora*.

Closes #1

Catches all `psycopg2` errors for now. And returns False
if a error is caught. Not a permanent solution, but works for now.

Haven't pushed to pypi yet. Will push after reviews.

`bcrypt.hashpw` returns the password hashed with the same salt,
not if the password matches the hash. So to check if the
password is correct, the hash thus obtained should then be
compared, and then verified.

Also, diaspora just doesn't hash the password using bcrypt, it
appends a "pepper" to the password and then hashes it. So, when
checking the password, the pepper should be appended.

`port` was supposed to be `db_port`
`user_id` is `@localpart:hs_bare` not `localpart:hs_bare`

It should work now.

For the time being, it authenticates against the diaspora db.
Many improvements are needed. Like no exception handlers
are registered.

TODO: implement exception handlers.

PyPi renders text in reStructuredText so it would be better to
have the README in reStructuredText than Markdown.