Doc comments cite a paper sentence that does not exist in it

A doc comment at `rs/cyphr-server/tests/witness_mode.rs:2046-2050` cites the factoring-trust paper and puts a sentence in quotation marks that does not appear in it:

> the split-view/equivocation analysis in `nrdxp/factoring-trust:docs/paper/_08-instances.qmd` **("a sequence cannot bind a history")** for why the test constructs a genuine two-branch equivocation rather than asserting the property abstractly.

Searching the paper's tracked sources for that phrase returns nothing. The word "sequence" appears in that file once, in an unrelated sentence about the mechanized chain being an ordered sequence of events. The paper does not discuss sequence numbers as anchors anywhere.

The sentence belongs to the Zami issue cited in the same breath, which is cited correctly. The quotation marks are what turned it into an attribution to the paper.

**The reasoning the comment gives is sound — only the anchor is wrong.** The paper does support the point, in the same file, where it treats the cross-verifier consistency claim:

> Posed over the joint record of shown views, this claim is determined and certifiable — a consistency proof is literally its certificate — and non-monotone: two views consistent today can diverge tomorrow, and the mechanization exhibits exactly that flip (over a two-element alphabet, which such a divergence provably requires).

That parenthesis is the actual justification for constructing a genuine two-branch equivocation rather than asserting non-monotonicity abstractly: the divergence provably requires two branches, and the mechanization exhibits the flip rather than stipulating it. The test's two independently-built chains are the direct analogue.

So the repair is to re-anchor the citation at that passage and drop the invented quotation, not to remove the reference.

Two related points found alongside it, both worth folding into the same change:

The second citation in the same file, at `:2261-2268`, is accurate — its quoted sentence is verbatim in the source. But its gloss is worth checking against what the paper actually types that claim as: over the canonical record the claim is determined, monotone and certifiable, and the trust is spent on *admitting the anchor* rather than on a monotonicity failure. The paper treats the anchor as an entailment move, which is a distinct notion from the axis failure the comment's framing suggests.

Both comments cite a file in a separate repository by path. That repository is public, so a resolvable URL would serve a reader better than a bare path, which reads as a file in this tree and is not one.

To reproduce:

```
git grep -n -i "bind a history"
git grep -n "witness_mode" -- rs/cyphr-server/tests/ | head
```

The first command returning nothing, against a comment that quotes the phrase, is the defect.

Start at rs/cyphr-server/tests/witness_mode.rs:2046-2050 and :2261-2268, then run the two git grep commands in the issue. Check the cited passages in nrdxp/factoring-trust:docs/paper/_08-instances.qmd, re-anchor the first comment without the invented quotation, review the second gloss, and replace bare paths with resolvable URLs. Done means both comments accurately describe their sources and the relevant test still passes.

Written by the indexing model from the issue text.