qcacld-3.0: Handle the peer leaks in sub system restart

When vdev destroy happens, vdev destroy command
is pushed into sme queue and driver waits for the sme
close session event to be completed, before this sme close
session is completed if SSR occurs, in SSR handling driver
completes all the events forcefully and hence vdev destroy
proceeds as the wait is completed because of SSR and
destroys the vdev logically. Now when this pending command
from the queue gets chance to execute, it tries to get the
vdev reference in wma vdev detach api to force cleanup of
the vdev peers, but as the vdev is logically deleted
because of force completion of the wait event during SSR
handling, it will not get the vdev reference and the vdev
peers will not be cleaned up. During stop modules this
peer leak will be detected.

To resolve this issue, force cleanup the vdev peers before
logically deeting the vdev when force completion of the
event happens because of SSR.

Change-Id: I88edf5028b818a499e1331f111202f6196aa8f03
CRs-Fixed: 2490486