cloud: Split deployment control into functions instead of services

At the moment, the deployment done with the `cloud` script can be split and controlled on Google Cloud service type boundaries (roughly). That was working quite well, but as the complexity of the deployment is growing, we need to consider splitting the deployment along functional boundaries. I.e. instead of "cloud functions", "pub sub", "postgresql", etc. split by function, such as "operational database", "archive database", "ingestion", "update processing", "notification generation", and so on.

Start by reading the `cloud` script and documenting how deployment is currently split across Google Cloud service types such as functions, Pub/Sub, and PostgreSQL. Define the functional boundaries described in the issue, including operational and archive databases, ingestion, update processing, and notification generation; the work is done when deployment can be controlled along those boundaries.

Written by the indexing model from the issue text.