MariaDB

[Iain-S]

Adds MariaDB support, which might also work for a wider array of DBs (sqlite perhaps?).

DSNs

This PR is a breaking change since we now require users to provide the DB connection DSNs (e.g. postgresql://user:secret@localhost:6789/dbname) rather than building them from SRC_DBNAME, SRC_HOST, SRC_PASSWORD, etc.
I find this less convenient, as an end user, but I think it's safest as the DSN is all we really need to pass to SQLAlchemy and this change

1. works for mariadb and postgres
2. future proofs us against other dbms we want to support
3. allows users to specify weird options, like this two-server postgres dsn (not tested) postgresql://host1:123,host2:456/somedb

Choosing providers

We modify the method of choosing a provider for a given column type. We now use a dict and try to .get() a match based on the column type and whether the column has a length attribute. If that fails, we iterate over all known column types and check issubclass(our_column_type, dict_column_type). This should result in fewer Unsupported type... runtime errors.

Loan tutorial

We add a short tutorial page to describe how to run SSG on this dataset.

SQLite3

Further to my speculation above, I have tested SSG with a simple SQLite3 DB using these env vars:

src_dsn="sqlite:////Users/me/somedir/src.sqlite"
src_dsn="sqlite:////Users/me/somedir/dst.sqlite"

and it works!

[cptanalatriste]

Looks good to me! Only minor re-naming suggestions.