fix(runtime): rewrite parentOperationId when copying conversations

[liuxiaocs7]

Summary

rewriteRuntimeEventReferences in conversation-copy.ts rewrote refs.operationId but let refs.parentOperationId fall into ...preserved, so it was carried into the copied session unchanged:

const { operationId: _operationId, traceEventId: _traceEventId, ...preserved } = event.refs;

This is always wrong after a copy, not sometimes. toolOperationIdMap rebuilds every operation id as buildToolOperationId({ invocationId: <target>, providerToolCallId }) under a freshly minted target invocation, and the write site (ai-sdk-backend.ts) sets parentOperationId: context.operationId — the same namespace. So a copied nested Code Mode call keeps pointing at the source session's operation while its parent has been renamed in the target. runtime-event-read-model.ts projects the field into StoredMessage, so parent/child pairing breaks, and tool-ledger-scanner.ts only shape-checks — the corruption is silent (like #3775).

The fix routes parentOperationId through the same rewriteOwnedId(..., references.operationIds, 'tool operation') as operationId, in the same destructure. The provider-owned parentToolCallId is deliberately left in ...preserved (it is not runtime-owned).

Fixes #3800

Verification

• Added conversation copy rewrites the parent operation id of a nested Code Mode call, which copies a Code Mode session and asserts the nested call's parentOperationId is rewritten to the target operation id while parentToolCallId is preserved.
• Confirmed the test fails without the fix (temporarily reverted the source change → 1 fail; restored → pass).
• packages/runtime typecheck (tsc) passes.
• biome check on both changed files passes.
• Full conversation-copy.test.js suite passes (20/20).

AI use

Select exactly one:

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

Tool(s) and scope: Claude Code — diagnosed the gap, implemented the one-line-scope fix in rewriteRuntimeEventReferences, and wrote the regression test. Human-reviewed.

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

[Astro-Han]

Update on de71830f59:

CODE GO — no P0-P3. Rewrites parentOperationId via runtime-owned operationIds map while preserving parentToolCallId; cross-run operation reuse is rejected, missing mapping fail-closes. Hosted test: SUCCESS (32859305478).

简体中文

该头无阻断。

[Astro-Han]

APPROVE — de71830, no P0-P3, hosted test SUCCESS.