[pr2eus] Add header timestamp to actiongoal topic in play-sound

[tkmtnt7000]

Now, when we execute something like (send *ri* :speak-jp "こんにちは"), sound action goal topic has no headertimestamp.
I fixed it in this PR.

# before this PR
$ rostopic echo /robotsound_jp/goal
header: 
  seq: 0
  stamp: 
    secs: 0
    nsecs:         0
  frame_id: ''
goal_id: 
  stamp: 
    secs: 0
    nsecs:         0
  id: "1686130902134708426_/fetch_eus_interface_1686130898541295371_2753725_robotsound_jp_0"
goal: 
  sound_request: 
    sound: -3
    command: 1
    volume: 1.0
    arg: "\u3053\u3053\u304Cdock\u3067\u3059"
    arg2: "ja"

[pazeshun]

In JointTrajectoryAction, timestamp of action goal message means when to start trajectory.
And if goal timestamp + time_from_start of some trajectory points are older than current time, these points are not executed.
This is sometimes troublesome especially when muliple PCs having independent clocks exist, so we sometimes feel setting timestamp 0 is better because it means immediate execution:
start-jsk/rtmros_common#1052 (comment)

I'm not sure about sound action (and timestamp added by this PR is action goal topic's one, not goal message's one), but please check carefully.

[k-okada]

@pazeshun I think ROS controller uses goal.trajectory.header.stamp, not goal.header

http://wiki.ros.org/pr2_controllers/Tutorials/Moving%20the%20arm%20using%20the%20Joint%20Trajectory%20Action#CA-11505f3470a9a076683bc46f9e917c9b05155c64_37

https://github.com/ros-controls/ros_controllers/blob/dce87855e1adf3a66271db8243d58561da9f2cdb/joint_trajectory_controller/include/joint_trajectory_controller/joint_trajectory_controller_impl.h#L610-L612

on pr2eus, we already set start time for both goal.trajectory.header.stamp and goal.header

jsk_pr2eus/pr2eus/robot-interface.l

Lines 838 to 853 in 8fc634d

	(send goal :header :stamp st)
	(cond
	((equal (class goal) control_msgs::SingleJointPositionActionGoal)
	(let* ((joint (car joints))
	(id (position joint (send robot :joint-list)))
	(pos (deg2rad (elt (elt (car trajpoints) 0) id)))
	)
	(send goal :goal :position pos)
	(send goal :goal :max_velocity 5)
	(send self :spin-once)
	)
	)
	(t
	(send goal :goal :trajectory :joint_names joint-names)
	(send goal :goal :trajectory :header :stamp st)

[tkmtnt7000]

I see. Thanks for your advice.

Now as long as I have checked, it seems that soundplay_node.py does not use timestamp of action goal topic, so I think this PR's change does not immediately have a side effect such as what you points out.
https://github.com/ros-drivers/audio_common/blob/3a748b223885a2f6035668c784e006e5426c84ef/sound_play/scripts/soundplay_node.py

FYI:
What I originally wanted to do was to avoid that the timestamp would be 1970/1/1 if the timestamp was left at 0 when using the function to remember the robot's speech and recall it later. (as part of our recent efforts around database_talker)