ratchet: re-pin igla race backend/rtl discard counts after #10689 moved one invariant #10695

#10689 moved the invariant rtl_count_mul_ops_comment_ignored_inv from specs/igla/race/rtl.t27 to specs/igla/race/backend.t27 to break their import cycle. The parser discards that invariant's forall ... ==> text in either file, so the same 15 tokens moved between two pinned parse-no-discard entries. The corpus ratchet on #10689 says:

• DISCARD WORSENED: specs/igla/race/backend.t27 discards 496 tokens, pinned at 481 (+15)
• DISCARD IMPROVED: specs/igla/race/rtl.t27 discards 613 tokens, pinned at 628 (-15)

The total did not change. The issue gate merged #10689 with this check red, so master's Corpus ratchet (expected-failure ledger) now fails for every PR. That is my miss: I compared test-report against master, but I did not run the ratchet before opening the PR.

Boundary

• In scope: re-pin the two entries in docs/reports/suite_expectations.json, with t27c suite --bless-expectations, the only writer (docs/CORPUS-RATCHET.md).
• Out of scope: making the invariant parse, the other igla race defects, and any other ledger entry.

User Scenarios

• A PR opened against master is not red on the corpus ratchet because of igla race: break the backend <-> rtl import cycle (Closes #10682) #10689.

Requirements

• The bless diff touches only the backend.t27 (481 -> 496) and rtl.t27 (628 -> 613) discard_tokens values, and their channel splits. The sum is unchanged.
• No entry is added, and max_entries does not rise.

Success Criteria

• On the Railway lab, t27c suite --repo-root . --ratchet --corpus-only on the PR branch reports RATCHET: PASS, or no DISCARD WORSENED line.
• The PR's own Corpus ratchet check is green.