Migrate to more robust database

[laurashub]

• Switch default database to MySQL for more robust handling of lots of data
• Provide method to convert existing SQLite databases to MySQL for legacy compatibility

[laurashub]

Ended up using MariaDB (forked from MySQL, so very compatible) since it has good support and is open source. I created a database named "hpp_db", and added a user 'hitpicker' with password 'hpppw'. My sql commands:

CREATE USER 'hitpicker'@'localhost' IDENTIFIED BY 'hpppw';
GRANT ALL PRIVILEGES ON 'hpp_db'.* TO 'hitpicker'@localhost IDENTIFIED BY 'hpppw';
FLUSH PRIVILIGES;

and to confirm its working as intended:

SHOW GRANTS FOR 'hitpicker'@localhost;
+------------------------------------------------------------------------------------------------------------------+
| Grants for hitpicker@localhost                                                                                   |
+------------------------------------------------------------------------------------------------------------------+
| GRANT USAGE ON *.* TO `hitpicker`@`localhost` IDENTIFIED BY PASSWORD '*501EE0D3242E85C0CF1887136077C78C166EDA49' |
| GRANT ALL PRIVILEGES ON `hpp_db`.* TO `hitpicker`@`localhost`                                                    |
+------------------------------------------------------------------------------------------------------------------+

In the config file (for production):

 SQLALCHEMY_DATABASE_URI = '{}://{}:{}@{}:{}/{}'.format(
        config( 'DB_ENGINE'   , default='mysql+pymysql'    ),
        config( 'DB_USERNAME' , default='hitpicker'       ),
        config( 'DB_PASS'     , default='hpppw'          ),
        config( 'DB_HOST'     , default='127.0.0.1'     ),
        config( 'DB_PORT'     , default=3306        ),
        config( 'DB_NAME'     , default='hpp_db' )
    )

The only change that I had to make to database.py (where the schema is defined) is the addition of the max number of characters allowed for strings. This isn't a huge problem, other than swapping things over from String to MEDIUMETEXT and LONGTEXT (ie saved PDB contents, smiles strings, mol blocks) since those can vary a lot and I'd prefer not to set a hard limit. The only fields where this is now relevant are:

• username (max 50 chars)
• screen name (max 50 chars)
• molecule name (max 50 chars)

Next step is making sure that container communication is working as intended since at the moment I'm testing this with Hitpicker running locally with no container while debugging. Additionally, before sending out this version I do want to make sure that existing SQLite databases with the previous schema can be ported over and will make sure to clarify the protocol with Helena to ensure it works as intended. The last thing I want is data loss.

I'm hoping that this swap will help the lag in dataloading Fang-Yu is experiencing since MariaDB is supposed to be faster, especially for large amounts of data and concurrent users. I'll also be implementing some of the changes mentioned in #10 which will help with that as well.