Fix .SortTokens() crash for states only seen inside a polymorphism

[ms609]

Summary

• .SortTokens() (R/Consistency.R) left mapping at 0 for any single state that only ever appears inside an ambiguous/polymorphic token (e.g. "(12)"), never on its own.
• wholes[x] then summed to 0 instead of a real combined code when remapping that ambiguous token, corrupting the effective state count and crashing ExpectedLength()'s downstream tabulate/sample pipeline (names<- length mismatch), which Consistency()'s rhi column depends on.
• Fix: after assigning codes to standalone ("whole") tokens, scan the full whole-token bit range for any position still unassigned and give it its own unused code, so ambiguous-token sums are always meaningful.

Repro:

library(TreeTools)
tr <- BalancedTree(paste0("t", 1:4))
dat <- StringToPhyDat("00(12)(12)", TipLabels(tr))
Consistency(dat, tr, nRelabel = 20)
# Error in names(object) <- nm : 'names' attribute [4] must be the same length as the vector [2]

Fixes #135

Test plan

• Added a direct .SortTokens() unit test for a state with no whole-token row
• Added an ExpectedLength()-level regression test reproducing the original crash
• testthat::test_file("tests/testthat/test-Consistency.R"): 25/25 pass
• spelling::spell_check_package(): clean
• GHA agent-check.yml on this branch: PASS

🤖 Generated with Claude Code