Add df.status_by_label() and clarify df.status() requires instance_id

[Copilot]

df.status() accepts an instance_id, not a label — passing a label silently returns NULL, making it easy for callers who use df.start(..., 'my-label') to unknowingly query the wrong identifier.

New function: df.status_by_label(label TEXT)

Returns the status of the most recently started instance matching the given label. Respects RLS (queries via SPI). Returns NULL when no visible instance matches.

SELECT df.start('SELECT 1', 'my-workflow');

-- Before: callers would mistakenly do this (always returns NULL)
SELECT df.status('my-workflow');

-- Now: correct label-based lookup
SELECT df.status_by_label('my-workflow');  -- 'Running', 'Completed', etc.

Labels are not unique — if multiple instances share a label, the most recently created one wins. Save the instance_id from df.start() when you need to track a specific run.

Changes

• src/monitoring.rs — status_by_label implementation; documents NULL-on-error behaviour (consistent with df.status())
• sql/pg_durable--0.1.1.sql — C wrapper registration for fresh installs
• sql/pg_durable--0.2.1--0.2.2.sql — CREATE OR REPLACE FUNCTION for the upgrade path
• docs/api-reference.md — warning note on df.status() clarifying it needs an instance_id; full entry for df.status_by_label()
• USER_GUIDE.md — updated Quick Status Check section with label-based example and tip
• docs/upgrade-testing.md — added df.status_by_label() to the monitoring functions table
• tests/e2e/sql/05_monitoring_and_explain.sql — tests correct status for a known label and NULL for an unknown one