run: PostRun detached ctx + panic-isolated Cleanup (gadfly #12)

Two convergent gadfly refinements on the PostRun wiring:
- PostRun now runs on detach(ctx), not the caller's ctx — a finished/cancelled
  caller no longer aborts artifact production (3-model: glm-5.2/minimax/deepseek).
- Cleanup is panic-isolated via safeCleanup (recover+log), matching runPostRun, so
  a misbehaving teardown can't clobber an otherwise-successful run (deepseek).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

run/executor.go

@@ -218,7 +218,7 @@ func (e *Executor) Run(ctx context.Context, ra RunnableAgent, inv tool.Invocatio
 	if inv.SessionToolFactory != nil {
 		st := inv.SessionToolFactory(&runSession{mailbox: mailbox})
 		if st.Cleanup != nil {
-			defer st.Cleanup()
+			defer safeCleanup(st.Cleanup) // panic-isolated, like runPostRun
 		}
 		for _, t := range st.Tools {
 			if err := toolbox.Add(t); err != nil {
@@ -338,7 +338,9 @@ func (e *Executor) Run(ctx context.Context, ra RunnableAgent, inv tool.Invocatio
 		if runRes != nil {
 			transcript = runRes.Messages
 		}
-		res.PostRunResult = runPostRun(ctx, postRun, transcript, res.Output, runErr)
+		// Detach from the caller's ctx: a finished/cancelled caller must not abort
+		// artifact production (the hook owns its own bounding, per its contract).
+		res.PostRunResult = runPostRun(detach(ctx), postRun, transcript, res.Output, runErr)
 	}
 
 	e.finishAudit(ctx, rec, status, res, started, runErr)