Reduce redundant network probing and prevent probing loop

NetworkMonitor did not limit the probe usage on a non-metered
network for data stall detection verification. As NM polls TCP
information with certain timer(10s for now), NM may get legacy
information if the data stall suspicion comes from other
signals, e.g. DNS. It may cause false alarm probing on the
network. The first tcp polling event will be 10s later. It may
cause a loop due to dns result received w/o latest tcp info.
Hence, start first polling while entering validated state.

Bug: 145275899
Bug: 147673885
Test: atest NetworkStackTests NetworkStackNextTests
Test: Manually test with such network and observe the probe
      behavior

Change-Id: Icb56ebe9d8304880d4f9f4fa8153b6e3727000fb