docs(cie): update SCHED_DEADLINE and re-application behavior

- Update yaml-specification.md: SCHED_DEADLINE is now applied
  immediately using SCHED_FLAG_RESET_ON_FORK instead of being
  deferred until all other CallbackGroups are configured
- Update integration-guide.md: document that the configurator keeps
  running and re-applies configurations when applications restart

Signed-off-by: atsushi421 <atsushi.yano.2@tier4.jp>

Author: atsushi421

docs/callback-isolated-executor/integration-guide.md

@@ -139,5 +139,7 @@ ros2 launch agnocast_cie_thread_configurator thread_configurator.launch.xml conf
 
 Then launch your application. The configurator applies scheduling parameters as CallbackGroups register.
 
+The configurator keeps running after all configurations have been applied. If the target application restarts, the configurator automatically re-applies configurations to newly discovered threads.
+
 !!! note
     The configurator automatically subscribes to all ROS domains referenced by `domain_id` in the YAML configuration. For the prerun mode, use the `domains` parameter to specify which domains to discover.

docs/callback-isolated-executor/yaml-specification.md

@@ -82,7 +82,7 @@ callback_groups:
     `SCHED_DEADLINE` uses cgroup cpuset for CPU affinity, which requires cgroup v1. See [Kernel boot parameter](integration-guide.md#kernel-boot-parameter-sched_deadline-only) for setup.
 
 !!! note
-    `SCHED_DEADLINE` configuration is applied after all other CallbackGroups have been configured.
+    `SCHED_DEADLINE` is applied immediately when the CallbackGroup is discovered using `SCHED_FLAG_RESET_ON_FORK`, which allows the thread to still call `fork(2)`/`clone(2)` after being placed under `SCHED_DEADLINE`.
 
 ## non_ros_threads