refactor(multi_object_tracker_node): isolate multiple measurement channels with dynamic callback groups

[kobayu858]

Description

Currently, the introduction of the Callback Isolated Executor into Autoware is underway.
To successfully achieve DAG (Directed Acyclic Graph) Scheduling via the Callback Isolated Executor, each callback recognized as a vertex in the DAG must belong to its own independent CallbackGroup.
This enables a one-to-one mapping between CallbackGroups and OS threads, allowing for fine-grained control over Priority and Affinity on a per-OS-thread basis.

Since isolating all callbacks at once involves a significantly high implementation cost, we are adopting a phased approach—starting the modifications and application with the major nodes that have the highest impact.

Detailed Changes

1. Dynamic Generation of Callback Groups Based on the Number of Channels

• Added cb_groups_measurement_ (a vector for holding the groups) to the header.
• Within the loop processing of the constructor, an independent MutuallyExclusive group is dynamically generated for each valid channel and assigned to each subscription via AUTOWARE_SUBSCRIPTION_OPTIONS.

2. Complete Protection of Internal State via std::mutex

• Newly added state_mutex_ to the header.
• Upon Data Reception (onMeasurement): To prevent interference from other threads during processing, the scope executing core::process_measurement and processObjects() is protected using std::lock_guard<std::mutex>.
• Upon Periodic Timer (onTimer): A lock is similarly applied to the scope handling transmission determination and distribution processing, completely preventing data rewriting (race conditions) during distribution.

Related links

Parent Issue: Separation of CallbackGroups: Background

• AutowareDiscussion about adopting CIE

How was this PR tested?

I ran lsim locally using a sample application.
I also measured the latency using CARET and confirmed that there were no significant differences.
The purpose of this PR is to enable scheduler configuration by isolating callback groups, not to reduce latency.

While the OSS version has only one subscriber, multiple subscribers are actually generated in practice.

Before

After

Notes for reviewers

• Ensure that individual callback groups are correctly and dynamically generated and applied to each channel.
• Verify that the scope of locks using state_mutex_ within onMeasurement and onTimer is appropriate, with no risk of deadlocks.
• Ensure that all accesses to shared resources (state_ / debugger_) are thoroughly protected due to multi-threading.

Interface changes

None.

Effects on system behavior

None.

[github-actions]

Thank you for contributing to the Autoware project!

🚧 If your pull request is in progress, switch it to draft mode.

Please ensure:

• You've checked our contribution guidelines.
• Your PR follows our pull request guidelines.
• All required CI checks pass before marking the PR ready for review.

[technolojin]

within the given mutex implementation context, odometry callback updateOdometryBuffer and it's buffer shall be mutex locked