Add LLVM coverage instrumentation workflow

[joce]

Overview

Introduce LLVM-based source coverage measurement for greenflame_core using Clang's instrumentation-based coverage (llvm-cov).

Changes

• CMakeLists.txt: Add GREENFLAME_ENABLE_COVERAGE option that applies -fprofile-instr-generate -fcoverage-mapping flags to greenflame_core when enabled
• tests/CMakeLists.txt: Apply coverage flags and link options to greenflame_tests when coverage is enabled
• CMakePresets.json: Add x64-coverage-clang preset with coverage instrumentation enabled
• scripts/coverage.ps1: New helper script that:
  • Configures and builds with coverage flags
  • Runs tests with LLVM_PROFILE_FILE set
  • Merges raw profiles with llvm-profdata
  • Generates summary and HTML reports scoped to src/greenflame_core/
  • Automatically adds LLVM tools to PATH from VS install
• docs/coverage.md: Comprehensive coverage guide covering prerequisites, workflow, implementation details, and manual steps
• docs/testing.md: Link to coverage documentation

Usage

.\scripts\coverage.ps1

Requires Clang compiler and LLVM tools from Visual Studio's "C++ Clang compiler for Windows" component.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c09ff35476

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Add VS LLVM bin when either coverage tool is missing

The PATH bootstrap only runs when llvm-cov is absent, but the script later requires both llvm-cov and llvm-profdata. On machines where llvm-cov is already on PATH (for example from another LLVM install) but llvm-profdata is missing or version-mismatched, this block skips adding Visual Studio's LLVM bin and the run fails or mixes incompatible tool versions. Gate this on both tools (or always prepend VS LLVM when VSINSTALLDIR is set) so the toolchain is consistent.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Prevent stale profraw reuse before merge

The merge step unconditionally consumes coverage.profraw without ensuring it was produced by the current test run. If a previous run left coverage.profraw behind and the current run crashes or exits before flushing profile data, llvm-profdata merge will still succeed on stale data and produce an incorrect coverage report for this invocation. Remove any existing raw profile before running tests (or verify fresh timestamp/existence after test execution) to avoid reporting outdated coverage.

Useful? React with 👍 / 👎.