coresight: move call to abort tracing after panic timeout delay

There appears to be a race condition in kernel panic where kernel sends
notification to all peripheral sub-systems as part of panic call backs to
prepare for graceful restart but before sub-systems respond to this
notification, we abort CoreSight trace. This results in losing info in
trace where sub-systems error handler is getting called. Move the trace
event to abort CoreSight tracing late in kernel panic after panic timeout
delay. It will give sub-system error handlers sufficient time to finish so
that CoreSight trace can capture this information.

Change-Id: Ib59de90dcb64f7a2a117a3886bd38359f3965130
Signed-off-by: Sarang Joshi <spjoshi@codeaurora.org>