Make fprime-util format work for libraries

[Sammy-Dabbas]

Fixes nasa/fprime#4926.

Problem

fprime-util format fails in an F Prime library (e.g. fprime-zephyr) for two stacked reasons:

1. format went through load_build(), which calls Build.find_nearest_parent_project(). A library has no root CMakeLists.txt with project(), so the command dies with "Could not detect project directory" before formatting ever runs.
2. The style file was hardcoded to framework_path/.clang-format from settings.ini, which libraries do not have.

Fix

• Add format to skip_build_loading() (alongside version-check) so no build/project context is required, and remove it from the now-irrelevant skip_build_cache_validation().
• New locate_clang_format_file() in commands.py: when a parent F Prime project is detected, use the framework's .clang-format exactly as before (no behavior change for deployments); otherwise walk up the directory tree from the working path for a .clang-format, mirroring clang-format's own discovery convention so libraries can carry their own style file.
• Guard builder.settings access in ClangFormatter.execute() so it tolerates build=None.

Testing

• 4 new tests covering the library cases: execute with no build object, style-file discovery from inside a library, the missing-style-file error path, and an end-to-end run_code_format against a synthetic library with its own .clang-format (skipped if clang-format is unavailable).
• Formatter suite: 11 passed (7 existing + 4 new), re-verified after rebasing onto current devel.
• Manually exercised the CLI against synthetic libraries: format and --check behave correctly, a missing .clang-format produces a clear error, and the deployment/project case still resolves the framework style file (regression-checked).

Happy to adjust the discovery order or error text if you have a preferred convention.

[Sammy-Dabbas]

Thanks. clang-format is already invoked with --style=file, so it discovers the nearest .clang-format from each input file's directory on its own. I dropped the manual walk up. The locate_clang_format_file now only resolves the framework's .clang-format when a project is detected and returns None for a standalone library, and the formatter skips its pre-flight existence check when there's no explicit framework file. The end-to-end library test still passes because clang-format finds the library's config. A library shipping no .clang-format falls back to clang-format's default style rather than erroring, which matches its normal behavior.

[thomas-bc]

Looking at this in a bit more detail, I'm realizing that things have been off for a while (nothing to do with your PR - so thanks for unearthing this!). The code goes and look for path/to/fprime/.clang-format, but it's never actually used.

Since #199, we only specify --style=file and never give an explicit path. So fprime-util format always delegates to clang-format to find the .clang-format file in a parent repo.

This is intended. Projects have their own .clang-format file, and libraries that want formatting should define their own as well. There's therefore no reason to try and find the fprime/.clang-format file. We should remove the logic to attempt to do it. Libraries can then run fprime-util format and it'll behave exactly the same: let clang-format try to find a .clang-format file in a parent directory.

[Sammy-Dabbas]

Done in 5e99471. The --verbose location print is left out since we no longer
locate the file ourselves. There's nothing for us to print, and clang-format
resolves it internally. Happy to add it back in some form if you'd like.

[Sammy-Dabbas]

Hi @thomas-bc, following up on this one. I pushed the changes you asked for; the framework .clang-format lookup is removed and discovery is delegated to clang-format via --style=file (5e99471, 6bac973). I left the --verbose found-location print out since the tool no longer resolves the file itself, but I am happy to add some form of it if you would like.