Use Absolute names for type hints

[InvincibleRMC]

Description

Should address the biggest problem of #257.
In the future I agree everything should be absolute but, for now makes the type hints introduced in #206 absolute to avoid global pollution.

Fixes # (issue)

Is this user-facing behavior change?

Did you use Generative AI?

Additional Information

[InvincibleRMC]

Pulls: #258
Gist: https://gist.githubusercontent.com/InvincibleRMC/56575dc8efb875212c7f1807908c7144/raw/e50b4e87196371b73468922d0896f8c6f051663f/ros2.repos
BUILD args: --continue-on-error --packages-above-and-dependencies rosidl_generator_py
TEST args: --packages-above rosidl_generator_py
ROS Distro: rolling
Job: ci_launcher
ci_launcher ran: https://ci.ros2.org/job/ci_launcher/18876

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

[InvincibleRMC]

Pulls: #258
Gist: https://gist.githubusercontent.com/InvincibleRMC/ac73e2ed0230b2c7e0dd0adf7f6517b7/raw/e50b4e87196371b73468922d0896f8c6f051663f/ros2.repos
BUILD args: --continue-on-error --packages-above-and-dependencies rosidl_generator_py
TEST args: --packages-above rosidl_generator_py
ROS Distro: rolling
Job: ci_launcher
ci_launcher ran: https://ci.ros2.org/job/ci_launcher/18881

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

Linux rebuild

[InvincibleRMC]

Pulls: #258
Gist: https://gist.githubusercontent.com/InvincibleRMC/92a492d7d78b193f13c777004c22f781/raw/45c0124fbc72901c8b826ac6a2e13c710633ed11/ros2.repos
BUILD args: --continue-on-error --packages-above-and-dependencies rosidl_generator_py
TEST args: --packages-above rosidl_generator_py
ROS Distro: rolling
Job: ci_launcher
ci_launcher ran: https://ci.ros2.org/job/ci_launcher/18922

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

Windows rebuild

[fujitatomoya]

lgtm with green CI.

@InvincibleRMC windows CI is still failing though... 👁️

[christophebedard]

I see a lot of the same failures in the last Windows release nightly: https://ci.ros2.org/view/nightly/job/nightly_win_rel/3763/#showFailuresLink

[christophebedard]

I checked and every single Windows test failure is in the Windows release nightly, and none of them seem really related to this PR. So I'll merge this.

[peci1]

Apparently, I'm late to the party. I'd just like to confirm that the most important behavior that resulted in wrong generated interfaces seems indeed fixed by this PR. There are still a few leftovers that pollute local namespaces, but they're not as important as those polluting the global namespace. Thanks!