Allow loading configuration from .env files.

[shtlrs]

Closes #227

The changes in this PR will now allow configuring pinnwand through a .env file.

The configuration in the .env file will not take precedence over its equivalent in the system's environment.

[shtlrs]

@supakeen I see the test_use_shorter_uri is failing because of the clean_bd fixture.

It's only used in one place, which seems a bit off since test_use_longer_uri is almost identical and isn't calling it.

Plus, why are you we clearing the tables if the database is destroyed later on along with its tempdirectory ? Is the point of it to ensure that the tables are empty before we start testing ?

I was thinking about completely deleting this, but wanted to double check first.

EDIT:
I wanted to test this even further, so I wanted to check the behavior on the master branch and I noticed the following:

For the same test instance, the database_uri in the database fixture is somehow different then the one in the clear_db fixture and the one that the main command sees before creating the tables.

This could be the reason why we have never seen this exception before, I'll have a deeper look later on to see if I can pinpoint why this is happening.

Second Edit:
After even more investigation, it seems this is a problem with the way the config is loaded, when I inspect configuration.database_uri and database._engine.url, the first has the path/database.db value and the engine is writing to the in-memory database. This seems because the import happens before we load the env variables.
I'll see how to fix this.

[supakeen]

Let's track these configuration issues in #232 for now. I'll take a look at these issues as well.

[shtlrs]

If #239 gets merged, I'll drop this one & make a new PR since the way we handle config will change.

[supakeen]

@shtlrs I've merged #239.

[shtlrs]

@supakeen Great! I'll make a new PR to deal with the related issue.