launch-upgrader: fix false completion detection (prompt contains the marker)

_completed() grepped the log for UPGRADE RUN COMPLETE, but the kickoff/resume
PROMPT (a user message) contains that string verbatim, so it false-positived
'done' while the run was still going. Check the model's ASSISTANT message output
via the web server API instead (log grep only as an offline, prompt-excluding
fallback).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

cc-ci-plan/launch-upgrader.py

@@ -245,9 +245,26 @@ def _run_pids():
     return out
 
 def _completed():
+    # The run is done only when the MODEL prints DONE_MARKER — i.e. it appears in an ASSISTANT
+    # message's output. NOT a log grep: the kickoff/resume PROMPT (a user message) also contains
+    # the marker (it instructs the model to print it), which would false-positive.
+    sid = _session_id()
+    msgs = _server_get(f"/session/{sid}/message") if sid else None
+    if msgs is not None:
+        msgs = msgs if isinstance(msgs, list) else msgs.get("data", [])
+        for m in msgs:
+            if ((m.get("info") or {}).get("role")) != "assistant":
+                continue
+            for part in (m.get("parts") or []):
+                t = part.get("text")
+                if isinstance(t, str) and DONE_MARKER in t:
+                    return True
+        return False
+    # Server unreachable → conservative log fallback that excludes the prompt's own mention.
     try:
         with open(LOG_FILE, errors="ignore") as f:
-            f.seek(0, 2); f.seek(max(0, f.tell() - 20000)); return DONE_MARKER in f.read()
+            f.seek(0, 2); f.seek(max(0, f.tell() - 20000))
+            return any(DONE_MARKER in ln and "print" not in ln and "'" not in ln for ln in f)
     except Exception:
         return False