timeout: add the --kill-after option

Based on a report from Kim Hansen who wanted to
send a KILL signal to the monitored command
when `timeout` itself received a termination signal.
Rather than changing such a signal into a KILL,
we provide the more general mechanism of sending
the KILL after the specified grace period.

* src/timeout.c (cleanup): If a non zero kill delay
is specified, (re)set the alarm to that delay, after
which a KILL signal will be sent to the process group.
(usage): Mention the new option.  Separate the description
of DURATION since it's now specified in 2 places.
Clarify that the duration is an integer.
(parse_duration): A new function refactored from main(),
since this logic is now called for two parameters.
(main): Parse the -k option.
* doc/coreutils.texi (timeout invocation): Describe the
new --kill-after option and use @display rather than
@table to show the duration suffixes.  Clarify that
a duration of 0 disables the associated timeout.
* tests/misc/timeout-parameters: Check invalid --kill-after.
* tests/misc/timeout: Check a valid --kill-after works.
* NEWS: Mention the new feature.

tests/misc/timeout

@@ -40,6 +40,12 @@ test $? = 2 || fail=1
 timeout 1 sleep 10
 test $? = 124 || fail=1
 
+# kill delay. Note once the initial timeout triggers,
+# the exit status will be 124 even if the command
+# exits on its own accord.
+timeout -s0 -k1 1 sleep 10
+test $? = 124 && fail=1
+
 # Ensure `timeout` is immune to parent's SIGCHLD handler
 # Use a subshell and an exec to work around a bug in FreeBSD 5.0 /bin/sh.
 (

tests/misc/timeout-parameters

@@ -35,6 +35,10 @@ test $? = 125 || fail=1
 timeout invalid sleep 0
 test $? = 125 || fail=1
 
+# invalid kill delay
+timeout --kill-after=invalid 1 sleep 0
+test $? = 125 || fail=1
+
 # invalid timeout suffix
 timeout 42D sleep 0
 test $? = 125 || fail=1