silx.util.retry: Fixed resetting the timeout timer when iterating

[woutdenolf]

This fix warrants a silx release. It affects all online workflows using blissdata.

[woutdenolf]

@t20100 @loichuder Fix for workflows that decide the scan ended while this is not true.

What happens is that the retry_timeout is reached (optional way to stop the data iteration in addition to the end_time presence). And it is reached because the timer did not get reset whenever data was yielded. Which means the timeout applied to the entire data processing instead of just the time it takes to get any new data. FYI the time it takes for a lima file to be completed is the time-limiting operation.