fix(nbd): change NBD status poll sleep from 100ns to 100µs to avoid useless busy spinning (#2884)

fix(nbd): adjust status poll sleep from 100ns to 100µs

Optimize NBD connected check polling interval:
- Original 100ns causes useless busy spin since nbdnl.Connect() is sync
netlink call;
- New 100µs yields CPU safely without any functional regression;
- Open/Close path have ctx & 30s timeout guard, rootfs start runs in
separate goroutine, no side effect for sandbox rootfs lifecycle.
/cc @jakubno @dobrac @ValentaTomas Would appreciate your review on this
change, thanks!

Co-authored-by: AdaAibaby <shaolila@buaa.edu.cn>
Co-authored-by: Jakub Novák <jakub@e2b.dev>

Author: 3 people

packages/orchestrator/pkg/sandbox/nbd/path_direct.go

@@ -232,7 +232,7 @@ func (d *DirectPathMount) Open(ctx context.Context) (retDeviceIndex uint32, err
 			break
 		}
 
-		time.Sleep(100 * time.Nanosecond)
+		time.Sleep(100 * time.Microsecond)
 	}
 
 	telemetry.ReportEvent(ctx, "connected to NBD")
@@ -325,7 +325,7 @@ func disconnectNBDWithTimeout(ctx context.Context, deviceIndex uint32, timeout t
 		if err == nil && !s.Connected {
 			break
 		}
-		time.Sleep(100 * time.Nanosecond)
+		time.Sleep(100 * time.Microsecond)
 	}
 
 	return nil