msm: adsprpc: do pm stay awake vote in rpmsg callback

Vote with PM to keep the CPU awake in the rpmsg interrupt callback
itself, instead of voting from the fastrpc thread after it is
woken up and scheduled, to handle race between actual  remote
subsystem response and context interruption because of CPU going
into suspend.

Change-Id: Ie97bf39e2ba0080cf738cc94cb65302acbd0f653
Acked-by: Thyagarajan Venkatanarayanan <venkatan@qti.qualcomm.com>
Acked-by: Maitreyi Gupta <maitreyi@qti.qualcomm.com>
Signed-off-by: Jeya R <jeyr@codeaurora.org>