Define, implement and document the migration process more specifically

[petteripitkanen]

Rename the migration process to database initialization process, and define that after each initialization the database will contain those crops and only those crops that are defined in the initial dataset. For convenience the initialization process will try to update the database when possible (to not change IDs, see #73). Those documents that are not in the initial dataset will be removed from the database.

When the database goes live the initialization process is not used anymore, instead there will be backups and possibly other specialized update processes.