Enable to call (one-shot-subscribe) safely in (let ...) scope

[708yamaguchi]

For detail, please see https://github.com/jsk-ros-pkg/jsk_roseus/issues/666

There are several cases where (one-shot-subscribe) occurs an error.

1. (one-shot-subscribe ...) -> C-c -> reset -> (let (a) (one-shot-subscribe ...)) ;; Bug
2. (one-shot-subscribe ... :unsubscribe nil) -> (let (a) (one-shot-subscribe ...)) ;; Bug
3. (one-shot-subscribe ...) -> C-c -> (one-shot-subscribe ...) ;; Bug
4. (ros::subscribe topic ...) -> (one-shot-subscribe topic) -> Bug
5. one-shot-subscribe -> C-c -> reset -> (ros::spin-once) -> Endless loop if a message is received
6. one-shot-subscribe -> C-c -> reset -> (let (a) (ros::spin-once)) -> Segmentation Fault if a message is received

In this PR,

• I solve problem 1. ~ 4. by removing :unsubscribe argument and checking subsciber. This makes sure that there is no subscriber before (while ... (ros::spin-once)) is called. (the first commit of this PR)
• I solve problem 5. ~ 6. by using class closures for callback function of (one-shot-subscribe). This allows the callback function to be accessed safely from anyscope. (the second commit of this PR)

My concern is that by removing the :unsubscribe argument, backward compatibility will be lost.

Thank you very much for your help. @YoheiKakiuchi @pazeshun @Affonso-Gui

[Affonso-Gui]

It would also be nice to have some more debate and some other points of view on the changes below:

• Removing the :unsubscribe argument. I am kind of comfortable with this change since I couldn't find any occurrences in the jsk_robot and pr2eus, but I'm not sure about hrpsys and tendon robots. If backwards compability is a must I guess we could also keep the key arg and just print a "deprecated" warning and unsubscribe it anyway?

• How to reject calls if prior subscribers are found. Most of the roseus functions log errors and return nil, but I think in this case it might be good to raise standard euslisp errors as well, since the program is most likely to briefly fail with something like cannot find method :... in ... when unsuccessful.

In some other minor comments:

• How about using ros::get-topic-subscriber instead of ros::get-num-publishers? This seems a lot more intuitive to me.
• The name of the second commit says "pointer closures" but I'm guessing you mean class closures?
• Some part of me would like to see the helper class as the internal symbol of some other package or with some underbar-ish crazy prefix that makes it very clear that it does not need to be directly invoked by the user.

[708yamaguchi]

Thank you very much for your review. I updated the code and pushed the commits.

I too would like to ask other people's ideas.

[Affonso-Gui]

@tkmtnt7000