'Period' should account for time loading and sending image

[benblamey]

The code should:

1. Figure out the timestamp when we need to send the next image next =(now + period)
2. Load, process, send the current one.
3. Sleep for (next - now)

This may help with issue when throughput is off for high frequencies

Update: doing this will not be enough to achieve theoretical max speed streaming - as a period of 0 results in 10 images per second. We need to read all the test images into memory, convert them appropriately, then send them. Theoretical maximum would be 50 images per second (for each thread), 3 threads needed to reach 150/second

[benblamey]

Under benchmarking, merely preparing the image for sending can take ~0.08secs, sending it takes ~0.02 seconds = total 0.1seconds per file. "Full speed" streaming will not exceed 10images/second

[benblamey]

38 frames / second x 4 color channels = 152/sec is theoretical maximum of the Yokogawa microscope

[benblamey]

With one processing container, processing is slower, so this issue does won't affect the overall throughput of the system.

[benblamey]

% See: benchmarking-simulator,2018-01-29-b Running the simulator... total time: 45.38119649887085secs durations as % of total streaming time: image_binning 0.0013759404926278853% image_to_bytes 64.32788427624905% prepare_image_bytes_method 68.11610430698585% read_image_from_disk 3.6551057999469525% stream_all_images 100.0% stream_file 20.374065210248805%