Flipped to explicit finite lock timeouts

With this version, all lock acquisitions will, by default, wait forever. For things like scan requests, a lock timeout is given explicitally to prevent accidental action queuing.

Reason for change: In the current release, if your storage IO is even a bit slow, scanning with raw data will fail because the IO takes so long, the fast autofocus thread can't acquire the camera lock. Rather than specify infinite timeouts at every point we want this, it seems more sensible to assume that, and specify instead the cases where we want to fail after a short time.

Assigned @rwb27 as a sanity check.