Document how to restart failed sync

[wonder-sk]

Sometimes users modify structure of their database and this naturally breaks db-sync. It would be good to document how to restart db-sync when this happens.

If the "source" is a geopackage in Mergin project (where the database structure change happened), this boils down to these steps:

1. delete "base" and "main" schemas in PostgreSQL
2. start mergin-db-sync again with --init-from-gkpg

If the source is PostgreSQL:

1. delete "base" schema in PostgreSQL and the geopackage in the Mergin project
2. start mergin-db-sync again with --init-from-db