Kill untracked process early

Untracked processes are processes whose process group is not lead by any
service. Such proceses can exist because, a process by default belongs
to the process group of its parent, but can decide join to a different
process group (via setpgid) and even become a new leader.

Previously, such untracked processes were killed (by sysrq) quite lately
in the shutdown sequence; after spending considerable amount of time
(5-6 sec) reaping those tracked processes. So, by the time when the
untracked processes are killed, we often are very close to the shutdown
timeout or already over it. Then init doesn't wait for these processes
to be completely reaped, and goes on to the next step of unmounting
partitions.

One consequence of this incomplete reaping is that the unmountings could
fail because some of the untracked processes may still be opening some
files on the partitions. init tries its best to do the unmounting by
trying it several times, hoping that the processes will eventually be
reaped. But when the timeout is over, init can't help but giving up and
shutting the filesystem forcibly down, which can manifest as a ramdump.

This change fixes this issue by killing those untracked processes early;
specifically along with those tracked services.

Bug: 433451565
Test: run a microdroid VM, and  `setprop sys.powerctl reboot`. Observe
kernel log:
[ 1278.258925] init: Received sys.powerctl='reboot' from pid: 2918 (setprop)
[ 1278.259826] init: Starting RebootMonitorThread
...
[ 1278.378265] init: Stopping untracked process group 2915 by sending SIGTERM
[ 1278.378416] init: Stopping untracked process group 2913 by sending SIGTERM
[ 1278.378526] init: Stopping untracked process group 2851 by sending SIGTERM
[ 1278.378692] init: Stopping untracked process group 2847 by sending SIGTERM
...
[ 1278.477114] init: Untracked process (pid: 2851 name: (binder:2851_2) ppid: 1 pgrp: 2851 state: Z) received SIGTERM
[ 1278.477219] init: Untracked process (pid: 2851 name: (binder:2851_2) ppid: 1 pgrp: 2851 state: Z) did not have an associated service entry and will not be reaped
[ 1278.514514] init: Untracked process (pid: 2874 name: (crosvm_VmRunApp) ppid: 1 pgrp: 2851 state: Z) received SIGTERM
[ 1278.514675] init: Untracked process (pid: 2874 name: (crosvm_VmRunApp) ppid: 1 pgrp: 2851 state: Z) did not have an associated service entry and will not be reaped

See the time difference between the start of the reboot thread and the
killing of the crosvm process. Previously it was 5-6 sec apart. Now it's
~100ms.
Flag: EXEMPT bug fix

Change-Id: Icb05171a71365acef0a692fe15ef78e62bc29d3b