fix: nested replacement tracked-change decisions

[harbournick]

No description provided.

[linear-code]

SD-3293

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 30467182c2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cubic-dev-ai]

cubic analysis

No issues found across 4 files

Linked issue analysis

Linked issue: SD-3293: BUG: Replacement inside another user's tracked insertion/deletion resolves parent and child changes incorrectly

Status	Acceptance criteria	Notes
✅	Create a single child replacement change (rather than separate insert/delete children) when a user replaces text inside another user's unresolved insertion.	Overlap compiler and tests were changed to expect one replacement child attached to the parent rather than separate insertion/deletion children; the test asserts change.type === Replacement and checks deleted/inserted segments and overlapParentId.
✅	Preserve the parent insertion/deletion as attributable to the original author (parent remains an independent change) when a nested replacement is created.	Compiler changes preserve the parent change and annotate child segments with overlapParentId; tests assert the parent change still exists and is of the expected type.
✅	Rejecting a nested replacement should restore/preserve the inferred parent deletion across split fragments (i.e., rejecting the child does not remove the parent deletion content improperly).	A new decision-engine unit test explicitly exercises rejecting a replacement that splits same-author deletion fragments and expects the parent deletion to be restored and remain as a single deletion change with the combined text.
⚠️	Accepting a nested replacement should resolve only the selected portion (replace only the selected part of the parent's inserted text) and keep the remaining parent insertion intact.	The decision-engine changes include logic to plan replacement resolution and the compiler creates a single replacement change, which are necessary to support correct acceptance behavior, but I don't see an explicit acceptance-resolution unit test in this diff that verifies accepting a nested replacement updates only the selected portion and preserves the rest of the parent insertion.
⚠️	Rejecting a nested replacement inside a parent's insertion should remove only the replacement and leave the parent's insertion intact.	The PR adds restoration logic for parent contexts and preserves parent metadata; however, the diff includes an explicit test for deletion-case restoration but I don't see a corresponding explicit test for rejecting a replacement inside a parent insertion in the provided changes.

_{Re-trigger cubic}

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!