Full clang-tidy gate runs one file at a time

[bburda]

Summary

The full clang-tidy gate (./scripts/test.sh tidy, i.e. the clang_tidy CTest test) analyses one translation unit at a time on any number of cores.

ament_clang_tidy defaults to --jobs 1, and our wrapper ros2_medkit_clang_tidy() in ros2_medkit_cmake/cmake/ROS2MedkitLinting.cmake passes neither a JOBS argument nor the ament_cmake_clang_tidy_JOBS variable that the upstream macro also honours. CTest-level parallelism does not compensate: each package registers a single clang_tidy test, so --ctest-args -j $(nproc) only overlaps packages, and the largest package still runs serially inside its own test.

Observed on a 16-core machine: the gateway package has 244 entries in its compile database, and a full run sits at 45+ minutes with a single clang-tidy process at 100% CPU while the other 15 cores idle.

Two paths are not affected and stay as they are:

• CI does not use this code path at all. quality.yml drives run-clang-tidy (parallel by default) with clang-tidy-cache, and analyses only changed files on pull requests.
• The pre-push incremental script scripts/clang-tidy-diff.sh loops serially, but pre-commit already invokes it with several file batches in parallel, so it is not the bottleneck.

So this is a local-developer cost only, paid every time someone runs the full gate before pushing.

Proposed solution

Pass the job count through the wrapper:

• add a JOBS argument to ros2_medkit_clang_tidy() and a ROS2_MEDKIT_CLANG_TIDY_JOBS cache variable defaulting to the host core count;
• run the tidy preset in scripts/test.sh with --ctest-args -j 1, because the parallelism now lives inside each package test - keeping both would multiply peak memory by the number of packages (7 packages register the test, and each clang-tidy process holds roughly half a gigabyte).

Additional context

The default stays conservative in the sense that it matches the machine it runs on, and a developer who wants a quieter box can set -DROS2_MEDKIT_CLANG_TIDY_JOBS=<n> at configure time.