Narrow R-parity claim to actual tested D set; align parity docstring

R0 review on the prior commit surfaced two actionable P3s:

1. The registry sentence "R-parity for D in {0..9} is asserted"
   overstated evidence: the cited scenario only exercises D in
   {0, 1, 2}. Narrow the claim to match what's actually tested and
   separately note that single-digit nonnegative D shares the same
   string encoding (no parity scenario currently runs outside
   {0, 1, 2}).

2. The TestDCDHDynRParityByPathNonBinary class docstring still
   used the narrower "breaks for D >= 10" wording. Align with the
   broader "multi-character baseline states (D >= 10 or negative D)"
   wording used in CHANGELOG and the registry.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: igerber