feat: add multi-database support via df.start() database parameter

[Pino de Candia (pinodeca)]

Summary

Allow durable functions to target any database on the cluster by adding an optional database parameter to df.start(). A single function invocation targets exactly one database; cross-database workflows can use dblink/postgres_fdw or separate durable functions.

Design spec: docs/multi-database.md

API

-- Existing (unchanged):
SELECT df.start(df.sql('SELECT 1'));
SELECT df.start(df.sql('SELECT 1'), 'my-label');

-- New: target a specific database
SELECT df.start(df.sql('SELECT 1'), database => 'analytics');
SELECT df.start(df.sql('SELECT 1'), 'my-label', 'analytics');

Invalid database names are rejected immediately at df.start() time.

Changes

File	Change
src/lib.rs	Add database TEXT column to df.nodes and df.instances DDL
src/dsl.rs	Add database param to df.start() with pg_database validation
src/types.rs	Add database to FunctionNode; add database param to connect_as_user()
src/activities/execute_sql.rs	Add database to ExecuteSqlInput, pass to connect_as_user()
src/activities/load_function_graph.rs	Include database in node SELECT
src/orchestrations/execute_function_graph.rs	Include node.database in SQL activity input
tests/e2e/sql/34_multi_database.sql	E2E test: cross-db execution, invalid db, regression
scripts/test-e2e-local.sh	Add test 34 to superuser list

Testing

• All 75 unit tests pass
• All E2E tests pass including new test 34 (cross-database execution, invalid database rejection, backward compatibility)
• Clippy clean, fmt clean

[Copilot]

Pull request overview

Adds multi-database targeting to pg_durable by allowing df.start() to specify a target database per durable-function invocation, while keeping pg_durable metadata and orchestration in the extension’s home database.

Changes:

• Extend df.start() with an optional database parameter and validate it against pg_database.
• Plumb the database value through nodes/instances metadata, graph loading, orchestration input, and SQL execution connections.
• Add an E2E test covering cross-database execution, invalid database rejection, and backward compatibility.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
src/dsl.rs	Adds database parameter to df.start(), validates existence, persists it to df.nodes/df.instances.
src/lib.rs	Adds database column to df.nodes and df.instances DDL; updates unit tests for new start() signature.
src/types.rs	Adds database to FunctionNode; extends connect_as_user() to accept an optional database.
src/activities/load_function_graph.rs	Loads database from df.nodes into FunctionNode.
src/orchestrations/execute_function_graph.rs	Passes node.database into execute_sql activity input JSON.
src/activities/execute_sql.rs	Includes database in activity input and uses it to connect to the right DB; improves trace logging.
tests/e2e/sql/34_multi_database.sql	New E2E test for cross-database execution + invalid DB rejection + regression behavior.
scripts/test-e2e-local.sh	Ensures test 34 runs under superuser (creates/drops a database).
docs/multi-database.md	Design spec documenting API, behavior, schema, and testing strategy.