Previously, the accounting data were built on demand,
extracting the data from the invoices on-the-fly.
This is intended to be used only once in a while, so there was
no performance issue with that.
Now, we want those data to be accessed from the OpenAPI,
so building them on-the-fly would be very much
intensive and resouces heavy. So we build them each nights
using a scheduled worker and save them in the database
