Add writer preparation evidence for Blog 6

[tom-clickhouse]

Blog 6's four writer-service visuals need downloadable SQL, measured source results and code that reproduces the published values. This adds an independent evidence package for 08a processing throughput, 08b ingestion CPU work, 08c compression and pre-aggregation reduction, and 08d background consolidation.

• Publish 97 retained SQL exports as 102 source CSVs, with a query/result index, schemas, hashes and explicit status for diagnostics without retained results. Preserve counters and timestamps while replacing hostnames with stable node aliases and omitting sampled execution IDs and repeated query text.
• Add Python standard-library derivation and export-conversion scripts, per-panel CSVs, exact four-visual payloads, shared CPU/memory utilization, intermediate measurements and accounting definitions. Document the active UTC window, smoothing, CPU sample allocation, merge-counter semantics and separation of later storage snapshots.
• Link the package from the existing writer utilization README. The package is independent of reader evidence PR Add reproducible reader metrics and query resource evidence for Blog 6 #44.

Validation: Python 3.9 regeneration and derive.py --check pass for all 35 derived artifacts. Every numeric value matches the approved 08a–08d payloads. All 102 source CSVs reproduce from the retained original SQL payloads. A deliberately changed output makes --check fail without modifying files. Source-publication checks, README links and git diff --check pass.

Evidence entry point: full-path-realtime/quotes/clickhouse-cloud/results_t2/utilization/writer/preparation-metrics/README.md.