Streaming cpu high usage fix

[abernste]

Bug fix: The CPU usage is too high when running Realsense Viewer while streaming is enabled.
Found the bug on post processing filters render loop thread
The thread is polling for new frames on a high frequency
The fix: change render loop thread from frame polling to wait for frame with small timeout (100 msec)
This will free the CPU (sleep) until frame arrives, without delay on new frames,
and not hang the render thread for more than 100 msec when video stream is disabled.

[abernste]

This fix is a suggested fix to: #1809

[abernste]

Wait for frames	CPU usage	Close delay
1	23%	Immediate
10	20%	Immediate
20	15%	Immediate
30	10%	Immediate
100	10%	Immediate
1000	10%	1 sec

I would go with 30 msec, agree?

[dorodnic]

I'm concerned that this will introduce frequent handled exceptions, preventing effective debugging of the application. The issue is with wait_for_frame throwing on time-out. Am I wrong?

[abernste]

@dorodnic
As I run this on release (to check CPU load), I didn't see any timeout exception.
Maybe on debug there are exception..
Maybe we can remove the timeout check in rs2_wait_for_frame on debug?
Maybe we can add a new default parameter to function single_consumer_queue:dequeue to ignore timeouts?
I think its better to use wait_for_frame from our framework instead of writing a new one just for the render function.
Do you suggest something else?

[dorodnic]

In #1730 we had a proposal to change the time-out behaviour of wait_for_frames. Overall, it would make sense. Lets discuss internally first.

[abernste]

Pending this pull request for wait_for_frame refactoring.
(Remove exception on wait_for_frame timeout)

[dorodnic]

Continued on #2055