Trusted evidence references must not be discovered recursively

Evaluator v2 review observation (2026-09-07, source: local ender-inference semantic.py): evidence-reference validation must use only server-generated reference IDs. Recursively accepting arbitrary ref keys lets untrusted prompt/schema/tool-result payloads manufacture citable evidence, undermining grounded LLM judgments. Preserve a separate trusted evidence index and fail closed for malformed or truncated evidence.