feat(desktop): preserve transcript continuity across bounded ranges

[Sun-GLiang]

Summary

Preserve transcript continuity while durable bodies remain bounded to the existing 512 KiB and 10-Turn range.

This adds a lightweight per-Turn position index, pages it through the existing Runtime Host transcript subscription, carries a bounded position sidecar through DesktopTranscriptReplica, and projects unloaded ranges as gap rows after the existing Turn materialization. Gap actions reuse the current loadBefore, loadAround, and loadLatest paths; scrolling remains owned by the existing scroll authority.

No full-history snapshots, generations, leases, second body protocol, byte scanner, or second transcript replica are introduced.

Fixes #4123

Verification

• Exact active-overlay Desktop E2E: while the final Turn is streaming, jump to the oversized first Turn, verify the durable gap, page through it, return to the tail, and complete overlay → durable handoff without duplicate/missing Turns or scroll jump (partial-history-notice.spec.ts, 2/2 repeated runs passed).
• 640-Turn two-sweep stress: mounted Turns max 10, position sidecar max 128, gap rows max 2; projected heap growth 4.39% and 6.41% (both below the 10% limit).
• Native seven-pair A/B sweep against PR base 6c8e749d3: median TaskDuration 883.385 ms → 880.742 ms (-0.30%); median RecalcStyleDuration 6.733 ms → 6.707 ms (-0.39%); candidate Long Animation Frames over 50 ms: 0/7; mounted Turns remained 10.
• Candidate spot run after the final navigation refactor: TaskDuration 815.973 ms, RecalcStyleDuration 5.286 ms, frame p95 18.4 ms, LoAF >50 ms 0.
• npm run typecheck, npm run lint, npm run format:check, npm run check:renderer-architecture, npm run check:asf-headers, npm run check:third-party-notices, and git diff --check passed.
• Full parallel npm test: 1,583 passed, 12 skipped, with two unrelated Runtime Host peer-mesh timing failures under concurrent load; the complete peer-mesh.test.js file then passed serially (20/20).

UI evidence

Active overlay beside a durable history gap:

Returned tail after overlay → durable settlement:

Review focus

• Schema v37 migration and bounded lazy position backfill
• Fixed-watermark and revision semantics for position cursors
• Desktop position-only retry and failure isolation
• Gap classification around active overlay handoff

AI use

Select exactly one:

• No generative tool made a substantive contribution
• Generative tooling made a substantive contribution

Tool(s) and scope: Codex implemented the storage, Runtime Host, Desktop, renderer, test, and review-fix changes. Each affected commit includes a Generated-by: Codex trailer.

Checklist

• Tests cover the change and fail without it
• Lint, format, typecheck and the affected suites pass locally

Does this PR entail a change in behavior?

• Yes — described under Summary above
• No