fix(sepia): define operative-with-priors in the report, align ledger and zh-CN README

Round 2 of Codex review on #46, three findings.

Mixed statuses within a family. A Claude Fable 5 role has one operative
table and three prior tables, and the report line allowed one value per
role. Rather than widen the schema, the value is defined: operative
means the release's own table is operative and the family's other
tables are priors. The definition sits on the report lines in
professional-pass.md and rubric.md and in the model-fingerprints.md
preamble, so a single status carries the whole family's state.

Ledger sentence. research/sources.md still said the consulted releases
"load nothing" — the same contradiction fixed in model-fingerprints.md
one round earlier, stated in a second place that was not opened. The
sentence now distinguishes families with no table (DeepSeek, Kimi: load
nothing) from releases inside a family with tables (Claude: no operative
row, family tables as priors).

Translation parity. README.zh-CN.md line 21 still carried the old
one-layer fingerprint sentence while README.md and README.zh-TW.md had
the two-layer version; CONTRIBUTING.md requires every maintained
translation to stay aligned with the English canonical. Updated in
Mainland terminology (测得、措辞层、应用、猜测).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TtoUEQarAYtQYrZLLwqjao