Revisit CFG->filelifetime implementation and uses

Doing some innocent testing of MDL-38743 it was discovered that CFG->filelifetime is not used as often as desirable.

See point B1 in that issue.

This is about to consider how we could be using it better and apply that new approach everywhere.

Perhaps it may have sense to move it to within the sendfile functions, so they continue accepting arbitrary lifetimes, but it CFG->filelifetime is set, it takes precedence. Or perhaps it only takes precedence if another settings says "force"… or only smaller accepted if debugging is enabled…

The trick here is to decide the behavior that should help both sites wanting to put a longer TTL and developers wanting to use small values for testing. While observing for it in as many places as possible (if set).

That is it, ciao