Move rmw_*_event_init() functions to rmw_implementation

[mm318]

Related to ros2/ros2#822

This moves the rmw_publisher_event_init() and rmw_subscription_event_init() functions into rmw_implementation so that they will be able to check for whether or not the given type of rmw_event_type_t is supported by the chosen middleware.

[mm318]

The full set of related pull requests are the following:

• Move rmw_*_event_init() functions to rmw_implementation #202
• Add rmw_*_event_init() functions to rmw_implementation rmw_implementation#88
• Add rmw_*_event_init() functions rmw_fastrtps#354
• Add rmw_*_event_init() functions rmw_connext#397
• Add rmw_*_event_init() functions rmw_cyclonedds#115
• Update docstring for rcl_*_event_init() functions rcl#600
• Add new type of exception to create_publisher/subscription() rclcpp#1026

(rclpy did not need any changes as far as I can tell.)

[mm318]

@ivanpauno and @wjwwood, can we kick off CI and see if these changes are ready to be merged?

[ivanpauno]

@ivanpauno and @wjwwood, can we kick off CI and see if these changes are ready to be merged?

Sure thing (try to provide a gist for this cases):

• Linux
• Linux-aarch64
• macOS
• Windows

[mm318]

Sure thing (try to provide a gist for this cases):

Sorry! Will do next time!

[mm318]

Does it look like we can merge this set of pull requests?

[ivanpauno]

I will rerun linux CI job that failed randomly:

• Linux

Most of the failures in Windows should been fixed after ros2/launch#391, the others seem random flaky tests failing. Re-running to check:

• Windows (unrelated failures)

[mm318]

@ivanpauno, it looks like the Linux build randomly failed again. Also, the test failures on Windows look unrelated.