Add event queue executor support.

[ZhenshengLee]

Brief feature description

Event queue executor was introduced recently by iRobot team as a strong mean to reduce the cpu overhead and latency by rclcpp layer.

See https://github.com/ros2/design/pull/305 for details.

AFAK, event queue executor + rmw_iceoryx + zenoh(mentioned in https://github.com/eclipse-iceoryx/iceoryx/issues/869) is much more performant than current default ros2 layers, both in latency and cpu usage.

So event queue executor would be implemented in rmw_iceoryx, perhaps in the rolling branch.

First of all, may galactic and rolling branch being created based on foxy branch?@Karsten1987 @budrus

Thanks.

[budrus]

@ZhenshengLee rolling branch could be the master branch, or? Indeed, we saw recently from measurements with G and zero-copy that for payloads < 64k the default executor has a higher impact on latency than the copies you have when not doing zero-copy

[ZhenshengLee]

@budrus

rolling branch could be the master branch

You are right, master branch can be used when foxy branch be merged into master, and galactic branch is needed anyway.

Indeed, we saw recently from measurements with G and zero-copy that for payloads < 64k the default executor has a higher impact on latency than the copies you have when not doing zero-copy

I've git it a test using APEX's performance_test tool(see custom version https://github.com/ZhenshengLee/performance_test/commit/c71666d04935572e634203ea951b7f5fa3cc5f8c), and cpu usage reduced about 40% in the test below.

./colcon/install/performance_test/lib/performance_test/perf_test -c rclcpp-events-executor --msg Array1k -p 1 -s 10 -l log_sub
./colcon/install/performance_test/lib/performance_test/perf_test -c rclcpp-single-threaded-executor --msg Array1k -p 1 -s 10 -l log_sub

Thanks.

[mossmaurice]

@ZhenshengLee I created the galactic branch and raised a PR. Will update master to work with rolling.