Add independent pipeline auto-tuning tool

[Toti330]

Summary

Adds tools/tune_pipeline.py, a standalone external benchmark orchestrator for worker threads, writer threads, and commit count.

Why this is useful

It replaces ad-hoc one-off environment changes with repeatable, randomized sweeps. Each candidate runs 3 times by default, reports min/median/max throughput, and recommends the smallest candidate within 95.00% of the best median. It also reports the efficient ceiling before performance falls below that threshold.

Implementation

• Supports independent --test worker, --test writer, and --test commit sweeps.
• Uses time.perf_counter() outside the application, avoiding application throughput-reporting artifacts.
• --dry-run executes 0 workloads and changes 0 files.
• --test-only sets TEST_ONLY=1: it requires an existing schema, bypasses schema/checkpoint work, and rolls each COPY transaction back so it persists 0 rows.
• Persistent writer/commit tests require --allow-db-writes and explicit --start-seed.
• --apply-to requests confirmation for every individual environment setting; rejecting one continues with the remaining recommendations.
• Includes 4 Python unit tests for recommendation, Docker argument expansion, environment-file updates, and partial confirmation.

Independence

This PR is based directly on main and can be reviewed or merged independently. It does not require the other diagnostics branches.

Verification

• python3 -m unittest -v test_tune_pipeline.py: 4 passed.
• python3 -m py_compile tune_pipeline.py test_tune_pipeline.py passed.
• cargo test --all-targets passed.
• Docker writer dry-run showed TEST_ONLY=1 and CHECKPOINT_FREQUENCY=none.

---

Generated by GPT-5.6 Terra & Sol. Ideas from the author; implementation by AI.