Replace std::default_random_engine with std::mt19937 (rolling)

[keeponoiro]

This PR addresses issue #2842, where std::default_random_engine can lead to periodic goal ID duplication due to its underlying implementation in GCC (minstd_rand0). When actions are executed frequently, this periodicity increases the likelihood of duplicate goal IDs, potentially causing unintended behavior.

In extreme cases, duplicate goal IDs can:

Trigger an exception when attempting to accept a new goal (goal ID already exists).

Extend the expiration time of the first goal when a duplicate ID is encountered.

To mitigate this, this PR replaces std::default_random_engine with std::mt19937, which provides a higher-quality random number generator with better distribution properties, reducing the risk of goal ID collisions.

Changes in this PR
std::default_random_engine → std::mt19937 in client_base.cpp

Ensures goal ID randomness to prevent collisions in high-frequency action executions.

Testing
In a 13-hour test, we observed periodic goal ID duplication when using std::default_random_engine.

Same goal ID appeared twice within a 10-minute period.

When goal ID retention was reduced to 1 minute, the issue no longer occurred.

With std::mt19937, goal ID duplication is effectively mitigated.

Impact
No ABI breakage.

Improves reliability of goal ID generation.

Addresses unintended goal expiration extension issues.

[jmachowinski]

Pulls: #2843
Gist: https://gist.githubusercontent.com/jmachowinski/fa696c2c5e21119166bc88323ebe7a7a/raw/09ab555f570f14b4ac797af1221c6430c7837dad/ros2.repos
BUILD args:
TEST args:
ROS Distro: rolling
Job: ci_launcher
ci_launcher ran: https://ci.ros2.org/job/ci_launcher/16015

• Linux
• Linux-aarch64
• Linux-rhel
• Windows

[jmachowinski]

@keeponoiro can you signoff your commit ?

[fujitatomoya]

i think this makes sense. lgtm with green CI.

• we do not need to let the platform decide the best engine here for the performance.
• consistent output across platforms with the same seed.

[keeponoiro]

@fujitatomoya
Sorry, I was referring to the situation in Humble. In Rolling, the goal timeout has been changed from starting at action initiation to being set at action completion, and the timeout itself has been reduced from 15 minutes to 10 seconds.

With a 10-second timeout, the probability of goal ID collisions is significantly lower. As @fujitatomoya mentioned, it would have been better to focus on maintaining consistent performance across platforms.

That said, since ROS 2 users can modify the timeout, having a lower probability of goal ID collisions when the timeout is extended is still beneficial.

[fujitatomoya]

@keeponoiro no worries. timeout for expiration is the really different topic, i was just saying that has been changed in jazzy and upstream distros.