claw: write the text helpers once in specs/claw/text.t27 instead of copying them #10317

Goal

Seven helper bodies are copied across specs/claw/ and sit in tools/duplicate_bodies_baseline.txt since #10312 (Closes #10311): buf_has, digit_byte, put_count, put_str (gates.t27 + sarif.t27), str_eq (gates.t27 + suggest.t27), has_prefix (service.t27) / starts_with (preview.t27), and the byte-equality loop same in sarif_pins.t27. Write each of them once, in one shared module, and delete the copies, so the ratchet goes down instead of being blessed up.

Boundary

• New file specs/claw/text.t27 (module ClawText) holding the counted-string helpers with their tests.
• specs/claw/gates.t27, sarif.t27, suggest.t27, service.t27, preview.t27, sarif_pins.t27: delete the local copies, add use claw::text;, pass the length as a parameter at each call site.
• gen/c/claw/gates.c regenerated and the seals whose spec text changed (ClawGates, ClawGatesPins, ClawService) re-saved, in the same commit.
• Not touched: tools/dupe_scan.py, the policy gate files, the behaviour of any helper.

Scenarios

1. t27c gen-c on a spec that uses claw::text returns 0 and the C output holds exactly the helpers the spec names (use_resolve splices them in before the backend runs). The comment in sarif.t27 saying the copies exist "because this file must lower alone" is wrong and is replaced.
2. t27c test-report on text, gates, sarif, sarif_pins, suggest, service, preview, github_surface and recipes: no failing test, none blocked.
3. python3 tools/dupe_scan.py is OK and the seven claw groups are gone from the known set.
4. The gate-topology claw drift step still compiles gen/c/claw/gates.c and prints "matrix is current"; tools/check_seal_currency.py reports 0 stale generated-code hashes.
5. preview.t27 keeps a single use: it imports claw::text only, because a second use with a same-named str_eq (recipes vs text) is reported ambiguous and the qualified recipes:: references stop resolving.

Requirements

• Every helper takes its string length as a parameter (gen-c cannot express .len).
• No behaviour change: the moved bodies are the existing ones; the callers differ only by the added length argument.
• No new hand-written code outside .t27.
• Seals and gen/c/claw/gates.c come from the lab build of the same tree.

Success Criteria

• dupe_scan reports the seven groups no longer present once --bless is run on top of dupe-ledger: bless the claw helper copies from #10283/#10286/#10287/#10288 (Closes #10311) #10312's baseline, and no new group.
• All nine claw specs above pass test-report on the Railway lab.
• t27c seal --verify MATCHes for every claw seal; check_seal_currency.py shows 0 stale.

Close criteria

Close when the PR that moves the helpers into specs/claw/text.t27 is merged and the dupe baseline no longer lists the seven groups.