Include source paths in launch errors

[le700]

Description

• Include the failing launch file path in InvalidLaunchFileError raised after all loaders fail.
• Add the included source location to missing-required-argument errors.
• Cover both diagnostics with regression tests.

This addresses the malformed-file and missing-argument cases described in #637.

Is this user-facing behavior change?

Launch loading and missing-argument errors now identify the source launch file.

Did you use Generative AI?

Yes. Codex assisted with code, tests, and PR text. The exact model identifier for the initial generation was not recorded. This follow-up uses GPT-6; GPT-6 Luna is used for independent review and verification.

Additional Information

• Focused load-error tests with ROS package imports stubbed: 3 passed.
• py_compile passed for all changed Python files.
• git diff --check passed.
• The full ROS pytest files could not be collected in this local environment because ament_index_python and lark are unavailable. CI should run the unmocked tests.

[christophebedard]

Please use the standard pull request template you get when you open a PR through the GitHub website. It also includes a section to declare whether you are using AI (and, if so, which model). Please read the full OSRA policy on use of AI: https://cuu.qwj.mybluehost.me/wp-content/uploads/2025/05/OSRF-Policy-on-the-Use-of-Generative-Tools-Generative-AI-in-Contributions.pdf