MINOR: debug: add "rt=1" to "debug dev task" to tune the RT flag

MINOR: debug: add "rt=1" to "debug dev task" to tune the RT flag

When rt=1 is set, the created task is set to real time. This will
essentially be used jointly with print=1 to print the wakeup date.

Now it is pretty much visible that TASK_RT helps keeping recurrent
delays stable under load:

  $ socat - /tmp/sock1 <<< "expert-mode on;debug dev sched task inter=1 count=100000"
  $ socat - /tmp/sock1 <<< "expert-mode on;debug dev sched task print=1 inter=1000 count=1"

  $ taskset -c 0  ./haproxy -db -f h2-h1.cfg
  [NOTICE]   (11262) : Automatically setting global.maxconn to 2033.
  task 0x5a67740: time_ms=355817139.2593
  task 0x5a67740: time_ms=355818165.934056
  task 0x5a67740: time_ms=355819192.609666
  task 0x5a67740: time_ms=355820219.528467
  task 0x5a67740: time_ms=355821245.778249
  task 0x5a67740: time_ms=355822271.800731
  task 0x5a67740: time_ms=355823297.836419
  task 0x5a67740: time_ms=355824327.894992
  task 0x5a67740: time_ms=355825354.92738
  task 0x5a67740: time_ms=355826381.242925
  ^C
  => ~1030ms interval

  $ socat - /tmp/sock1 <<< "expert-mode on;debug dev sched task inter=1 count=100000"
  $ socat - /tmp/sock1 <<< "expert-mode on;debug dev sched task print=1 inter=1000 count=1 rt=1"

  $ taskset -c 0  ./haproxy -db -f h2-h1.cfg
  [NOTICE]   (11274) : Automatically setting global.maxconn to 2033.
  task 0x1c89d740: time_ms=355842657.804670
  task 0x1c89d740: time_ms=355843657.130495
  task 0x1c89d740: time_ms=355844657.159396
  task 0x1c89d740: time_ms=355845657.176695
  task 0x1c89d740: time_ms=355846657.144914
  task 0x1c89d740: time_ms=355847657.416217
  task 0x1c89d740: time_ms=355848657.192072
  ^C
  => ~1000ms interval

Also mixing one non-rt and one rt tasks easily shows that the non-RT skews:

  task 0x3180080: time_ms=356074147.738657
  task 0x7f82705749c0: time_ms=356074167.140429
  task 0x3180080: time_ms=356075157.58365
  task 0x7f82705749c0: time_ms=356075167.155128
  task 0x3180080: time_ms=356076167.23508
  task 0x7f82705749c0: time_ms=356076167.163529
  task 0x7f82705749c0: time_ms=356077167.131421
  task 0x3180080: time_ms=356077176.204301
  task 0x7f82705749c0: time_ms=356078167.172336
  task 0x3180080: time_ms=356078186.76828
  task 0x7f82705749c0: time_ms=356079167.110302
  task 0x3180080: time_ms=356079195.814623

Under traffic with a run queue of 50k, with non-rt we can see delays
of up to a few hundred ms, while the rt one is stable:

  task 0xf52228e3c580: time_ms=12817821323.738291
  task 0xf52228e3c580: time_ms=12817822375.954267
  task 0xf52228e3c580: time_ms=12817823437.542732
  task 0xf522c85fe340: time_ms=12817823901.552931
  task 0xf52228e3c580: time_ms=12817824487.769129
  task 0xf522c85fe340: time_ms=12817824901.491252
  task 0xf52228e3c580: time_ms=12817825553.539234
  task 0xf522c85fe340: time_ms=12817825901.930177
  task 0xf52228e3c580: time_ms=12817826613.611365
  task 0xf522c85fe340: time_ms=12817826901.94851
  task 0xf52228e3c580: time_ms=12817827659.386720
  task 0xf522c85fe340: time_ms=12817827901.655777
  task 0xf52228e3c580: time_ms=12817828694.795097
  task 0xf522c85fe340: time_ms=12817828901.882341
  task 0xf52228e3c580: time_ms=12817829765.216886
  task 0xf522c85fe340: time_ms=12817829901.723780
  task 0xf52228e3c580: time_ms=12817830844.706481
  task 0xf522c85fe340: time_ms=12817830901.571820
  task 0xf522c85fe340: time_ms=12817831901.440380
  task 0xf52228e3c580: time_ms=12817831951.556375

This indicates that thisshould be a sufficient first step to unblock
haload.