docs: fix stale v2.0 documentation and remove LLM artefacts

[ms609]

Audit of the documentation surface touched by the 2.0.0 release, checking each claim against the current implementation. Findings were gathered by clustered read-only passes; every fix below is backed by the code it contradicts.

Stale / incorrect claims

• large preset docs described a preset that no longer exists. large is now an alias of thorough (presets$large <- presets$thorough) with maxReplicates raised to 500, but the roxygen, a SearchControl() comment, and search-algorithm.Rmd / tree-search.Rmd still described the retired bespoke preset (biased Wagner, simulated annealing, reduced cycles).
• default preset documented as "12 ratchet"; it is 6 (ratchetCycles = 6L).
• Vignette preset claims contradicted SearchControl() defaults: adaptiveStart "disabled in large" (it is TRUE), ratchetTaper / annealCycles "enabled in thorough/large" (both off), consensusStableReps "default 3" (default 0), and the auto-selection tip/pattern thresholds.
• Profile parsimony "binary only (max 2 states)" in MaximizeParsimony() — multi-state is supported (Carter 1990 for 2 states, Maddison & Slatkin 1991 for 3–5, Monte Carlo beyond). ConcordantInformation() carried the same stale two-state restriction.
• PrepareDataProfile(): documented a split.sizes bootstrap entry it never writes, and claimed simplification happens "with a warning" when only the inapplicable-token case emits a message.
• WideSample() @return said "uniform sample"; it returns a Max-Min diverse subset (the function's whole point).
• fractional-weight default documented as 1000L; it is 1260L.
• Implied-weight fit formula rendered as k / e + k; corrected to k / (e + k) (propagates to inheriting man pages).

LLM artefacts removed

• The "reflects island structure, not a cap" defensive framing in the Completeness section. (The "256 is not hard-coded anywhere" aside that prompted this audit was already removed on the branch tip by Replace Morphy with native kernel for inapplicable data #251.)
• NEWS.md: the "This addresses a common complaint…" defensive aside, the empty ## Other improvements section, and a duplicated PaintCharacters() bullet.

Deliberately left for the maintainer

• The # To integrate into 2.0.0 notes scaffold in NEWS.md now also holds Replace Morphy with native kernel for inapplicable data #251's breaking-change bullets (MorphyLib removal, function renames), so how it folds into the released 2.0.0 section is that PR's call, not this one's. I applied only the unambiguous fixes within it.
• The **renamed to \Morphy()`** line in the 2.0.0 breaking-changes list is now stale, since #251 removed Morphy()` — flagging rather than rewriting, as the correct wording depends on Replace Morphy with native kernel for inapplicable data #251's intended migration story.

man/*.Rd regenerated with roxygen2 8.0.0.

🤖 Generated with Claude Code