ci(test): parallelize coverage and hosted diagnostics

[gbaudrit]

Summary

Restores the functional CI critical path close to its pre-coverage behavior by running uninstrumented tests in build-and-test, collecting functional coverage on two balanced parallel runners, and keeping hosted Management memory diagnostics off the critical path.

Changes

• Moves the five sequential hosted module memory measurements from build-and-test to a dedicated parallel hosted-module-diagnostics job while preserving budgets and artifacts.
• Runs the required Fast and Integration functional lanes without instrumentation in build-and-test.
• Splits coverage collection into two balanced solution inventories, merges their 19 raw Cobertura files, and publishes one unchanged functional-code-coverage artifact.
• Keeps performance projects, test assemblies, and generated sources outside the functional report.
• Documents the parallel coverage execution model and retains the single-command local coverage workflow.

Validation

• Pre-coverage GitHub baseline: build-and-test completed in 6 min 45 s; the preceding unsharded coverage run took 9 min 34 s.
• GitHub Actions run 34610662062 passed: build-and-test completed in 5 min 52 s, improving the unsharded coverage run by 3 min 42 s (39%) and the pre-coverage baseline by 53 s.
• Coverage shard A passed in 5 min 22 s; shard B passed in 4 min 17 s; consolidation and artifact publication passed in 42 s.
• The full CI critical path completed in 6 min 18 s and is now bounded by hosted-module-diagnostics, not coverage.
• Local coverage shard A passed: 139 Fast tests and 221 Integration tests (219 passed, 2 optional integrations skipped).
• Local coverage shard B passed: 471 Integration tests (470 passed, 1 optional PostgreSQL integration skipped).
• Local consolidation passed for 19 module reports: 16,890 / 26,427 lines (63.91%) and 7,187 / 11,546 branches (62.25%); test and performance sources were excluded.
• All pull request checks passed and the pull request is mergeable.
• Trade-off: parallel coverage duplicates runner setup and compilation, increasing aggregate runner minutes while reducing wall-clock latency.
• Desktop and mobile smoke testing not applicable; this change affects CI orchestration only.

Breaking changes

None.