t27c run-record: --receipts <dir> and --json <path>, so the Chip tab shows t27c's own judgment #7723

Part of epic #6655 (Verified Compute). The Spec Explorer's Chip tab (gHashTag/trinity#955) shows a spec's hardware record, but only hand-transcribed figures for hand-written RTL. Since R2/R3, Verilog compiled from a spec has run on the die with signed receipts that t27c run-record judges (#7598: ternary_link, complete, FRESH, Die NAMED). The website cannot run t27c, so it needs that judgment as data, produced by t27c itself, never re-implemented in TypeScript.

Goal

t27c run-record can judge a run kept in any directory and write its judgment as JSON.

Boundary

• bootstrap/src/main.rs: two options on RunRecord and its two dispatch sites.
• bootstrap/src/service.rs: run_run_record glue only.
• bootstrap/tests/signed_receipt_reader.rs: one test.
• No spec change. The decisions stay in run_record.t27, signed_receipt.t27, die_binding.t27 and independence.t27.
• Text output unchanged.
• Inside the own-language budget: at most 40 hand lines per file, 80 per diff.

User Scenarios & Testing

• Given three receipts in .trinity/receipts/run1/, when run-record --receipts .trinity/receipts/run1 --json out.json runs, then out.json holds:
  • run_complete true, authentication FRESH, die NAMED;
  • an independence line starting INDEP_DIES;
  • three receipts;
  • judged_by naming the t27c build.

Requirements

• FR-1: --receipts <dir> (default .trinity/receipts) selects the receipt directory; a run archived in a subdirectory can be judged.
• FR-2: --json <path> writes the judgment just printed. Every value is one already computed; nothing is re-judged.

Success Criteria

• cargo test --release -p t27c --test signed_receipt_reader passes, including json_writes_the_judgment_of_a_run_kept_in_a_receipts_subdirectory.
• A mutant that ignores --receipts fails that test.
• lefthook run pre-push passes (own-language, foreign-budget).