msm: rpm: panic() on an rpm error interrupt

If the RPM triggers the LOW interrupt it means the RPM has error
fataled. Handle this case so we don't sit spinning on the RPM to
respond until our watchdog times out. We also check this while
polling the ack interrupt in msm_rpm_busy_wait_for_request_completion().

This should allow us to have a nice panic output in the logs
whenever the RPM has crashed and is no longer responsive.

Change-Id: Ia40b95eaffe5bbaa10b62f85ae849eb36b1384e6
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>