[HanielUlises]

At the end of a mission, the launch file interrupts the demo performers. About one run in five then segfaulted after every action had finished (2 of 10 standalone interrupts). gdb puts the fault in the performer's destructor, run from main after rclcpp::shutdown(): CascadeLifecycleNode frees its activators_state_ map through a pointer that is no longer valid. No thread of the process is using the node at that point, only CycloneDDS's own.

The three demo performers now end through eplansys_demo::leave(), which shuts ROS down, flushes stdout and stderr, and leaves through _exit, skipping a teardown that has nothing left to do. plansys2_tests exits the same way for its own exit-time fault. The fault itself is in cascade_lifecycle, and fixing it there is a separate job.

Checked locally:

• Interrupts: 30 of 30 standalone interrupts exit 0. Before the change, 2 of 10 exited 139.
• Full demo: the two-site survey, serial and parallel, ends with every process finishing cleanly.
• Lint: ament_cpplint and ament_uncrustify pass on the package.