Subsystem failure should cause runtime to stop

There's a bit of a problem with state management when the storage sub-system fails.

A typical caller acquisition loop probably looks something like:

    while( cpx_get_state(runtime)==DeviceState_Running && ... )

When a write fails, the storage thread will exit, but the runtime fails to monitor that failure and stays in a "Running" state.

[ ] the subsystem failure should cause the runtime to do a controlled stop and we should be able to reliably query the system state.
[ ] There's a testing gap around failing devices.

acquire-project / acquire-common