* fix: advance postgres replication slot lsn via periodic standby status updates
* test: add e2e regression test for postgres replication slot lsn advancement
Drives sustained change traffic so the slot freeze reproduces deterministically (fails pre-fix at the 20s deadline, passes post-fix within ~10s). Also wires the postgres_trigger feature through windmill-test-utils and the integration-tests crate so the postgres trigger e2e tests are actually runnable.