cgwatch.sh respawns ratlock/cgdetect within ~30s if either dies (crash,
OOM, accidental kill), so the guard self-heals instead of staying down
until the next reboot. service.sh now launches the supervisor after the
RIL-up wait rather than starting the two daemons directly.
action.sh: use short pgrep patterns so the enforcement/detector 'running'
check resolves correctly instead of falsely reporting 'not running'.
cgdetect.sh: wrap dumpsys telephony.registry in timeout so a blocking
registry read can no longer freeze the status file and stall the frame
counter.
