fix(mcp): read the successes once, so every failed run is checked against all of them (#5842)

still_failing was handed the query's one-pass result. The first failed
run's any() consumed it, so a second failure in the window found no
successes left and was reported though a later run had put it right
(ci-runner-volumes came back once keepalived-survey also had a failure).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

steward/mcp/tools/overview.py

@@ -191,7 +191,8 @@ def still_failing(failed, succeeded) -> list:
     """The failed runs nothing has since put right: a later success of the
     same playbook from the same source on the same hosts clears one, as long
     as it was at least as real (a dry run's success clears only a dry run's
-    failure)."""
+    failure). `succeeded` may be a one-pass result, so it is read once."""
+    succeeded = list(succeeded)
     return [f for f in failed if not any(
         _run_key(s) == _run_key(f) and s.started_at > f.started_at and (_dry(f) or not _dry(s))
         for s in succeeded)]