docs(sync): record the b24ui-only useTokenSearch divergence as a porting invariant

[IgorShevchik]

Linked issue

Refs #339, #347, #363.

Type of change

• Documentation (updates to the documentation or readme)
• Bug fix (a non-breaking change that fixes an issue)
• Enhancement (improving an existing functionality)
• New feature (a non-breaking change that adds functionality)
• Chore (updates to the build process or auxiliary tools and libraries)
• Breaking change (fix or feature that would cause existing functionality to change)

Description

Revised after review. The first revision also carried a bullet marked PENDING, describing a truncation-units contract from an unmerged PR. It has been dropped: it stated a failure mode the loop cannot produce, its self-deletion instruction told the porter to edit a file outside src/ — which the same document forbids — and nothing enforced its expiry. Two factual claims have also been corrected; see the end of this description.

Documentation only. .sync/PORTING.md gains one §2 invariant plus a dated changelog line; .sync/nuxt-ui.json gets a one-line correction. No src/ change.

The divergence. highlight() takes a fifth parameter useTokenSearch and carries the token-search logic around it. c502157b added tokens / minTokenLength, 6743f793 added the parameter itself and gated the tokenizer behind it — both against the file's old name src/runtime/utils/fuse.ts, both shipped in v2.8.0. Immediately before c502157b the function took four parameters and computed no minTokenLength at all; that is what establishes it as locally authored.

Why it stayed invisible. The port 557a5178 renamed fuse.ts to search.ts and carried the divergence across. A pickaxe on the current path returns exactly one commit — 557a5178, which is a genuine upstream port with an Upstream: trailer — so the divergence is not merely hidden but actively misattributed. --follow is needed to see the two commits that introduced it.

Why it matters. Replaying upstream's four-parameter signature wholesale drops behaviour shipped in v2.8.0, and nothing goes red: useTokenSearch appears nowhere under test/. #363 tracks that gap.

Ledger correction. .sync/nuxt-ui.json's summary for 2a172ef asserted "highlight signature matches 1:1" without qualification. The .md log for the same port was already accurate — it scoped its claim to behaviour "without token search" — but the machine-readable half was not, and a porter consulting it would have got the wrong answer.

This follows the convention used by prior rule-adding PRs (#330, #331, #335, #338, #341, #343, #346, #351): a rule in the relevant section plus a dated line in "Changelog of rules". No processed entry — that map is keyed by 40-hex upstream SHA, and there is no upstream SHA here.

Corrections from review of the first revision:

• "the one invariant in this section with no guard at all" — false, removed. Locales, playground-manifest mirroring, jsDoc and Types are unguarded too. The text now says plainly that some §2 rules name a guarding spec and some do not, and points at test(search): highlight()'s b24ui-only useTokenSearch argument has no coverage #363.
• "neither carries an Upstream: trailer" as evidence of local authorship — weak, reframed. Only 16 of ~3200 commits carry that trailer, and 559a5cdb — this very file's most recent upstream port — is not one of them. The claim now rests on the four-parameter signature preceding c502157b.
• The quotation from .sync/log/2a172ef….md was truncated mid-parenthetical with a closing paren supplied that the source does not have. Now quoted in full.
• The upstream assertion is hedged in the file itself, not only in the commit message: nuxt/ui has not been re-inspected, so "upstream has no such parameter" is an inference from b24ui's own history.
• Attribution: the "byte-identical with upstream" premise originated here, in 595923b9 (PR fix(CommandPalette): stop the raw label and suffix reaching v-html #338), was repeated in bug(search): truncateHTMLFromStart splits astral characters, emitting lone surrogates #339, and was inherited in good faith by the external contributor whose PR prompted the check. The changelog line records that.

Checklist

• I have linked an issue or discussion.
• I have updated the documentation accordingly.

Note on scope: the branch name is left over from an unrelated task earlier in the session and does not describe this change.