[Feature] Improve coverage reliability and cover reader and commit edge cases #506

Search before asking

• I searched existing issues and found no issue tracking these reader/commit regression cases and concurrent coverage counters. Related report filtering work is tracked in [Bug] Coverage report includes system headers, third-party dependencies, and test sources #488.

Motivation

The coverage report exposes untested reader and commit paths: empty parent paths, default prefetch ranges, stream forwarding and I/O metrics, discarded batches, deferred read ranges, workloop errors, recovered commits, schema-dependent statistics, and row-ID conflict checks.

Coverage collection also needs to handle concurrent execution reliably. Non-atomic gcov updates can lose counter increments when reader threads execute the same code. The coverage script currently omits several optional components from its build configuration.

Solution

• Add 12 focused regression tests for local files, prefetch readers, snapshot commits, and conflict detection.
• Enable atomic gcov counter updates for C and C++ coverage builds and remove that flag from bundled third-party dependencies.
• Enable Jindo, OSS, S3, Lumina, Lucene, full-text, and REST components alongside the formats already enabled by the coverage script.

Anything else?

Implementation: #505. No public API, storage format, or protocol changes.

Focused suites passed 212 tests with 2 existing skips. Coverage flags, concurrent counter updates, third-party flag isolation, and the report script's command flow were checked separately. The full coverage report has not been regenerated: rebuilding the shared configuration is blocked by its stable Rust 1.67.1 toolchain being older than Vortex dependency requirements.

Are you willing to submit a PR?

• I'm willing to submit a PR! See test: improve coverage collection and cover storage edge cases #505.