Adopt sqlx migrate (#1221)

We now use `sqlx` ([1]) and its CLI ([2]) to manage our SQL schema and
migrate it to new versions. The schema is now represented as a series of
SQL scripts in `db/`, along with teardown scripts to enable rollbacks.
The objective here is to enable easy migrations, namely:

 - adding new tables
 - adding indices to tables
 - adding columns with default values to tables

More complex migrations which transform or delete data may require more
sophisticated tools than what's here, and even these "easy" migrations
require some planning and operational tools that aren't present in this
repository.

This commit contains these changes:

 - the `publish-sql-schema` workflow now publishes all migration scripts
   in `db` to a Google Cloud Storage bucket to later be consumed by
   Divvi Up ops tools
 - the interop test containers use `sqlx::migrate::Migrator` instead of
   `janus_cli` to write the schema into their database
 - `janus_aggregator_core::datastore::test_util::ephemeral_dataastore`
   uses `sqlx::migrate::Migrator` to apply migrations to ephemeral
   databases
 - `db/schema.sql` is copied to a `sqlx` migration script, and a
   corresponding `down` script is added which deletes tables, indices,
   etc.

[1]: https://crates.io/crates/sqlx
[2]: https://crates.io/crates/sqlx-cli

Part of #30

.github/workflows/publish-sql-schema.yml

@@ -29,4 +29,7 @@ jobs:
       id: get_version
       run: echo VERSION=${GITHUB_REF/refs\/tags\//} >> $GITHUB_OUTPUT
     - name: "Upload schema file(s)"
-      run: gcloud alpha storage cp db/schema.sql gs://janus-artifacts-sql-schemas/${{ steps.get_version.outputs.VERSION }}/schema.sql
+      run: |-
+        gcloud alpha storage cp --recursive \
+          db \
+          gs://janus-artifacts-sql-schemas/${{ steps.get_version.outputs.VERSION }}/