Superuser route to flush the metrics collector now

POST /observability/flush handles what the collector has queued and writes
it, so the soak harness no longer sleeps FLUSH_INTERVAL_S before clearing
its own rows.

backend/scripts/soak.py

@@ -36,7 +36,6 @@ from typing import Any
 import httpx
 
 from fluksio.core.config import settings
-from fluksio.flow.metrics import FLUSH_INTERVAL_S
 
 log = logging.getLogger("soak")
 
@@ -374,6 +373,14 @@ def create_flow(
     log.info("created flow '%s'", name)
 
 
+def flush_metrics(client: httpx.Client) -> None:
+    """Have the metrics collector write what it holds now."""
+    try:
+        client.post("/api/v1/observability/flush").raise_for_status()
+    except httpx.HTTPError as exc:
+        log.error("could not flush the metrics collector: %s", exc)
+
+
 def delete_flows(client: httpx.Client) -> None:
     for name in FLOWS:
         try:
@@ -887,11 +894,11 @@ def cleanup(client: httpx.Client | None) -> None:
         delete_flows(client)
 
     clear_redis()
-    # Deleting a flow publishes an audit event of its own, and the metrics
-    # collector buffers what it hears for up to FLUSH_INTERVAL_S before writing
-    # it. Clearing the tables any sooner races that flush and leaves the rows
-    # this very cleanup caused.
-    time.sleep(FLUSH_INTERVAL_S + 5.0)
+    # Deleting a flow publishes an audit event of its own, which the metrics
+    # collector holds until its next flush. Clearing the tables before it is
+    # written would leave the rows this very cleanup caused.
+    if client is not None:
+        flush_metrics(client)
     clear_observability()