Align sync-server record and CRDT transports

[h4yfans]

What

Align the sync server with the adapter-driven client sync model by separating record-sync and CRDT transport handling, preserving legacy record endpoints, and adding per-domain sync telemetry.

Why

Phase 07 requires the server transport to match the new client sync architecture. The existing sync route mixed record and CRDT validation, batching, quota handling, and logging in one place, which made the transport boundary harder to reason about and harder to observe per domain.

How

• grouped record transport behind shared handlers and added /sync/records/* aliases while keeping the legacy /sync/push, /sync/pull, /sync/changes, /sync/manifest, and /sync/status routes working
• moved record batch intake into processRecordPushBatch and reused shared payload hydration helpers in the sync service instead of keeping that orchestration in the route layer
• added SyncTelemetry structured logs for accepted/rejected record mutations, replay/conflict/quota rejects, per-domain record traffic, CRDT update/snapshot volume, and latency buckets
• extended sync route tests to cover record transport aliases and CRDT endpoint separation

Type

• feat — new feature
• fix — bug fix
• refactor — restructure without behavior change
• style — visual/UI only
• perf — performance improvement
• test — adding or updating tests
• chore — tooling, deps, config
• docs — documentation only
• ci — CI/CD changes

Test plan

• Unit tests added/updated

• Integration tests added/updated

• Manual testing (describe below)

• pnpm --filter @memry/sync-server test -- src/routes/sync.test.ts src/services/sync.test.ts

• pnpm typecheck:sync-server

Checklist

• Self-reviewed the diff
• No hardcoded secrets or credentials
• Files stay under ~500 LOC
• Follows immutable data patterns