msm: kgsl: Process HFI message queue inline

When we send out a command to GMU, we do an
unrealistic wait for five seconds for processing
the ack from GMU. This code path is invoked with
device mutex held, and therefore this wait can starve
other kgsl threads leading to fence timeouts and
performance hit. So reduce the timeout to a
reasonable scale. Also, so far this queue is used for
acknowledgements of commands sent to the gmu. Since
these acks are synchronous, we can process them inline
instead of using interrupts and scheduling a tasklet.
This way, We also get rid of unnecessary lists, their
locks and gmu interrupts.

Change-Id: Id6724d6e4ce489d776816d68933a2a8d30fd4ece
Signed-off-by: Harshdeep Dhatt <hdhatt@codeaurora.org>