cpufreq: flush queued work before freeing policy

When system resume and suspend happening at the same time,
the resume sequence will schedule the work (policy->update)
before policy->update getting chance to run in the meantime
suspend or hotplug of an cpu happens at the same time resume
is happening, and if the cpu is the last in the cluster
policy is freed.

This triggers an issue when scheduler tries to schedule the work
which is scheduled in resume path But the handles are freed already.

Flush the Queued work if any before freeing policy.

Change-Id: I41504e7d9c201916457681537ac480c92e7cbe5d
Signed-off-by: Santosh Mardi <gsantosh@codeaurora.org>