[Draft] Merging the upstream/rolling EventsExecutor here

[jfinken]

Getting this started. This PR pulls in the EventsExecutor code from the upstream rclcpp, rolling branch into here.
Locally:

• the unit-tests passed
• the examples run (including with the irobot_lock_free_events_queue (which remains where it is...)
• I modified the iRobot ROS2-performance-framework to utilize this branch, and thus this EventsExecutor.

With respect to the ros2-performance framework, on an iRobot robot, I ran 60-sec white_mountain multiple times with each EventsExecutor:

• the latency results are the same across the two EventsExecutors
• however the CPU-utilization results are rather different

EventsExecutor from the dedicated iRobot repo, CPU-utilization, last line

61000          7.1       62812          58253          20828          81972          2578636        

"upstream" EventsExecutor, from the experimental namespace, CPU-utilization, last line

61000          9.4       62644          58278          20828          82116          2578488        

• second column is the percent-util
• 7.1% vs 9.4%, a 32% increase in CPU