feat(migration): migrate historical review data Cloudflare D1/KV → self-host, then deprecate cloud infra

[JSONbored]

Priority: P3 (after reviews are flawless). Child of #1029.

Problem: self-host review_audit/review_targets/repository_settings are EMPTY — no historical reviews/outcomes/per-repo configs migrated from the cloud (Cloudflare D1 "gittensory-api" + KV). We cannot safely delete cloud infra until 100% migrated + verified.

Deliverables: (1) inventory the cloud D1 tables + KV namespaces (review history, pr_outcomes, repository_settings, per-repo rules/configs); (2) an export→import path into the self-host SQLite with row-count + checksum verification; (3) replicate each repo's REAL cloud config into its container-private .gittensory.yml (currently generic advisory templates); (4) a deprecation checklist for the old Cloudflare Workers/D1/KV/queues once verified. Analyze extensively before any deletion.

[JSONbored]

Inventory done (read-only); plan ready; NOT executed (risk gated). Source = D1 gittensory (b2c79dd6) ONLY: review_audit 126258 (awesome-claude 97467 / metagraphed 25773 / gittensory 3018), review_targets 3233, repository_settings 3, orb_pr_outcomes 481, pull_request_reviews 4221, gate_outcomes 236. Per-repo configs = the 3 repository_settings rows (NOT KV — both review KV namespaces empty). reviewbot D1 + KV REVIEW_CONFIG + KV METAGRAPH_CONTROL = stale/empty → deprecatable. ⚠️ CRITICAL: cloud repository_settings carry autonomy=full-AUTO + ai_review_mode=block — a verbatim import would flip the advisory self-host engine into LIVE auto-merge/auto-close. Import with autonomy/mode NEUTRALIZED to advisory. Self-host 6 core tables empty (bulk-load) but advisories/audit_events/check_summaries/pull_requests have native rows → INSERT OR IGNORE. Cloud is HOT → freeze cron/webhook or tail-sync before cutover. GUARDRAIL: delete nothing until the App webhook is redirected (needs #1572 + #1570) + a rollback window. Export scripts staged in scripts/export-d1-*.mjs.