msm: npu: refactor wait_for_status_ready to reduce the latency

This change is to refactor function wait_for_status_ready to
support register polling without context switch. If the fw
processing time is very short and the task requires short
processing latency, polling mode should be used.

Change-Id: Ia6cd630451eee593f6e07ce73172fac4def8b9d8
Signed-off-by: Jilai Wang <jilaiw@codeaurora.org>