Migrate coverage pipeline to score_coverage 0.3.1

[dcalavrezo-qorix]

Rollout of the qualified S-CORE coverage tool, score_coverage 0.3.1 (release notes), tracked in eclipse-score/score#3258. Linux LLVM coverage only; QNX on-target coverage is not part of this PR.

What

The coverage pipeline this repository used from @score_tooling//coverage moved to its own module. This PR switches the labels one-to-one (defs.bzl, merger, generate_coverage_html, the death-test feature) and adds score_coverage 0.3.0 as a dev dependency. rules_rust is already at the required 0.68.2-score. Scope (quality/coverage/BUILD), the coverage:llvm_cov config, the justification file and the workflow are otherwise unchanged; the README points at the tool's user manual.

Verification

Local, bazel coverage --config=llvm_cov //score/... --build_tests_only (37 tests), old pipeline vs this branch, LCOV compared record by record:

Check	Result
Files in report	77 → 147 (70 added, see below)
Covered/uncovered status of the 77 common files	identical for 75; 2 explained below
Raw line coverage	77.29 % → 78.35 %
Raw branch coverage	65.70 % → 67.40 %
HTML index	147 links, 0 dead, one row per file
Second run with the new module	identical status in every file
bazel build --nobuild --lockfile_mode=error --config=x86_64-linux //...	pass

Added files. 12 .h and 58 .hpp headers of health_monitor and launch_manager. The old pipeline dropped headers reached through include directories (eclipse-score/coverage_tool#5, fixed in 0.2.0); they are now measured. Most of them are well covered, which is why both percentages go up. 5 of the 70 are at 0 %.

Status differences.

• process_group_manager/details/process_info_node.hpp lines 61–72: the member-initializer lines of the move constructor are now reported (all covered); the old report showed only the braces of that constructor.
• osal/details/posix/semaphore.cpp line 100: the baseline run recorded 528 calls to wait() of which one sem_wait failed, the new runs 527 calls with none failing. A timing effect of the concurrency tests, not the tool.

New in the archive. unmapped_files.txt lists in-scope files that have no coverage data at all: 18 declaration-only headers and 33 no-data files. The no-data list includes Rust sources of health_monitor (lib.rs, the mod.rs files, log.rs, protected_memory.rs, supervisor_api_client) and launch_manager/src/alive/src/rust/src/lib.rs. None of these had coverage data before either; they were simply not listed. Worth a look by the module owners.

Update: bumped to 0.3.1, which requires the explicit --instrumentation_filter=^//score[/:] this branch sets. Bazel guesses the filter from the packages of the test targets; libraries tested from a test subpackage were otherwise compiled without counters or, on QNX, lost their counters (score/os showed 13 % instead of 80 % in the first QNX report). See the 0.3.1 release notes.