Support passing keyword arguments to rosidl CLI extensions

[hidmic]

This patch builds on top of #567, adding support for passing keyword arguments for CLI extensions along with their spec. This allow for extension specific configuration.

CI up rosidl_cli, rosidl_typesupport_c, rosidl_typesupport_cpp, and rosidl_generator_py:

• Linux
• Linux-aarch64
• macOS
• Windows

[hidmic]

@clalancette @sloretz friendly ping

[clalancette]

I don't think I have enough context to review this. Two questions that might help clarify things a bit:

1. How is this intended to be used in the end? That is, what is a realistic example of something that might get passed through here?
2. Do we really have to define our own new string format for the SPECS_PATTERN? Can't we use straight YAML (or XML, or JSON, or....)?

[hidmic]

@clalancette fair questions.

How is this intended to be used in the end? That is, what is a realistic example of something that might get passed through here?

It really depends on each generator implementation. A realistic example, the one that pushed me to do this, is narrowing down the pool of typesupport implementations to be used when generating CPython extensions and C/C++ typesupport trampolines. Simply looking up the ament_index assumes the corresponding code has been generated and built, which is true in the typical rosidl CMake pipeline (where presence in the index implies ament extension registration) but not when manually recreating a subset of it in a different build system :)

Do we really have to define our own new string format for the SPECS_PATTERN? Can't we use straight YAML (or XML, or JSON, or....)?

It is mostly YAML. We could go from py[key: value] to {name: py, kwargs: {}} plus py being a shortcut to {name: py}. I don't have a strong opinion.

[hidmic]

@clalancette @sloretz friendly ping

[hidmic]

Re-running CI:

• Linux
• Linux-aarch64
• macOS
• Windows

[hidmic]

Alright, thanks for the reviews! I'll get this in and release so we can have the Rolling PR job passing for downstream PRs.

[hidmic]

@cottsay how do you feel about backporting this (along with ros2/rosidl_typesupport#112 and ros2/rosidl_python#133) to Galactic? It is a feature and it does change the expected extension API but it is hardly likely anyone has used it outside of the core.