Introduce PROCESSINFO_TRIGGERMODE_SEMAPHORE_PROP_TIMEOUTS and make it…

[DasVinch]

… default for milk-streamFITSlog.

Adding triggermode 5, which is also a semaphore mode.

In triggermode PROCESSINFO_TRIGGERMODE_SEMAPHORE (==3), the TIMEOUT is caught and the overall computation is shunted into a continue;
In PROCESSINFO_TRIGGERMODE_SEMAPHORE_PROP_TIMEOUTS (==5), the timeout is willfully propagated, the execution of the loop iteration begins, and the timeout status can be checked in processinfo if need be. This actually emulates the old semaphore behavior when timeouts would propagate (by default every second) and click the AO chain unit after unit.

This fixes the following issues in logshim:

• FITS cube is finalized and dumped upon stream timeout
• FITS cube is finalized and dumped if saveON is set to False after the last frame

@oguyon foresee any pitfall?

[DasVinch]

FYI this branch is currently live on scexao5, not ported anywhere else yet.