Fix the probe wake lock duration when mcast_resolicit is enabled.

When mcast_resolicit parameter is enabled, kernel will fall back to
send multicast NSs numResolicit times if numSolicit unicast probe
(re)transimissions fail(i.e. never receive any confirmation from the
neighbor). In this case, wake lock duration should be updated by adding
the extra duration of multicast NS (re)transmission. Otherwise, we
might get an incorrect nud failure event from isFromProbe function.

For example, consider the below case:
1. NUD probe is triggered by L2 roaming;
2. send unicast NS probe by numSolicits times;
3. fall back to send multicast NS probes by numResolicits times;
4. NUD probe fails(no response);
5. isFromProbe() returns false due to the duration since the last
   timestamp of calling probeAll is greater than wake lock duration;
6. NudEventType will be considered as a organic failure instead of
   post roaming failure. However, it was triggered by L2 roaming.

Bug: 162944199
Test: atest NetworkStackTests NetworkStackIntegrationTests
Change-Id: I63098238eea98540edcbca58727efa2ed4e2ca36