fix(read): hide a revision nested inside a hidden one in the accepted/baseline views

[kklimuk]

🤖 docx-cli triage

What was wrong

The reader decided whether a run is visible in --accepted/--baseline from its innermost tracked-change wrapper only. A deletion nested inside another author's insertion therefore showed its text in read --baseline, even though rejecting every change removes it:

$ docx read doc.docx --baseline      # Editor's <w:del> nested in Reviewer A's <w:ins>
The Client disputed amount           # wrong — should be "The Client "

What changed

• One rule, one place: a run is visible in a view only if every enclosing tracked change is visible in that view (isRevisionVisible in src/core/track-changes/revision-visibility.ts). Accepted hides deletions/move-sources, baseline hides insertions/move-targets, current shows all.
• The reader records the enclosing chain on nested changes only (TrackedChange.within, optional; the run's own kind/id stays the innermost wrapper). read --ast is unchanged for documents without nested tracked changes.
• All view consumers use the predicate: read (markdown), find, wc, iterateBlocks({ view }). track-changes list output shape is unchanged. Schema widened for within.
• Replace/edit engines and accept/reject machinery are untouched.

Verification

• Now read --baseline ≡ track-changes reject --all and read --accepted ≡ accept --all on: a deletion nested in another author's insertion, a third author's deletion nested in an insertion, and an insertion nested in a deletion. Tests assert all six equivalences plus wc counts and find --baseline no longer matching nested-deleted text.
• Golden comparison vs main: read (all views, --ast), wc (all views) and find (all views) over all 36 fixtures — 540 commands, 0 differences.
• bun run check ✅ bun run test:unit ✅ (1736 tests, +12 new, no existing expectation changed).
• Independent /code-review xhigh of the full diff: no regressions against main. Pre-existing gaps it noted, left alone on purpose (each would change read output for existing fixtures or is outside this change): the markdown renderer still prints a <w:br/>/<w:tab/> inside a hidden wrapper; tracked-deleted table rows are hidden by read but not by find/wc; equations and note references ignore views; revN grouping is nesting-blind (fixed separately in the replace --track inside another author's pending insertion credits the new text to that author, and misses the target next to a nested deletion #13 PR).

Decisions for you

• None required. If you want the pre-existing markdown break/tab gap fixed, say so and I'll file it as an issue.

Part of the fix for #13 (the minimal #13 PR handles the write side; this handles the read side).

🤖 Generated with Claude Code

https://claude.ai/code/session_012GoxD36rnNfg6uArkVxtJL

---

Generated by Claude Code