Draft: Improve FVP-based test runs by enabling parallel execution
FVP is useful for exercising CPU features when dedicated hardware is unavailable, but it is slow and can significantly increase CI wall-clock time. To mitigate this, run FVP-backed tests in parallel and add support tooling:
- parallelize test execution and shard workloads in .gitlab-ci.yml
- add parallel.py to orchestrate test sharding and round-robin/bin/time-balanced strategies
- add fvp.sh wrapper to simplify running binaries inside the FVP environment
- fetch and reuse previous artifacts to avoid re-running already-completed work
- reduce long job timeouts and enable job-level parallelism where appropriate
This reduces overall CI latency while preserving deterministic coverage of SME/SVE/other feature paths.
Split huge test suites into smaller for better balancing.
Signed-off-by: Anton Bondarenko anton.bondarenko@arm.com