refs #386

refs #386

refs #386

refs #386

refs #386

refs #386

refs #386

Fix #385

refs #385

refs #385

Refs #385

Since this is what it indexes now

last_played_date already has its own dedicated index, so do not index it
a 2nd time.
This also allows ordering by insertion_date to leverage the index, while
it wouldn't have before since the colum was the second one.
refs #385

refs #385

refs #385

Fix #383

Which will be helpfull each time we need to generate a migration test
database

And add a way to specify how many artists we want

This version contains the sqlite3 CLI tool

More jobs will soon generate more files

Instead of having a different variable for each manually triggered job
(which would grow exponentially as we add more manual jobs) rely on a
common variable that hold the name of the job

Refs #372

Fix #372

refs #372

Since an album track is denoted by the media.subtype column, we will
need to monitor this field as well to update the nb_tracks field, and
can't assume that an album track removal will lead to a media deletion.
The subtype is more likely to be updated.

It's also clearer to have a dedicated trigger for a dedicated task

refs #372

Fix #381

The old version was only ever used to set a subtype to unknown in a
buffered way.
This simplify the function to always set the subtype to unknown and
doesn't buffer the changes

Which does not buffer the change and does not trigger a refresh
refs #381

refs #381