HSLM release close-out: clear master's inherited reds (scan, duplicate-bodies, orphans) so #10836 lands #10935

Epic #10024 (own model). HSLM release close-out. The release itself is done: #10467 merged as b2a33e5, the checkpoint is public at Hugging Face playra/hslm-t27-2m commit 39df9a77794988a0fff324b2b52fbe9176271a76, and PR #10836 (cell 25 published, Closes #10833) carries the reviewer bee's review 5567974497, pinned in specs/igla/own_model.t27 at head 4af190d. What is left is master's inherited red checks, which fail every PR that merges master, #10836 included.

Boundary

specs/tri/worktrees/reap.t27 and its seal; the tools/dupe_scan.py ledger rows for specs/igla/coder/hslm_bwd_ptx.t27 (or a shared helper spec that removes the copies); the bootstrap orphan ceiling. No Rust edits: if a mod line in Rust is needed, file an issue and stop. Out of bounds: specs/igla/own_model.t27, specs/hf/model_card_hslm.t27, specs/ml/train_hslm.t27.

Scenarios

• scan ("Block hardcoded developer home paths") fails: tools/check_devhome_paths.py reports FAIL new specs/tri/worktrees/reap.t27 (1 occurrence), a test string near line 253 that starts with a developer home path (added by tri worktrees --reap-plan: a REMOVABLE verdict for merged, clean lane trees, ruled in .t27 (Closes #10805) #10832 / ledger: the five rows master lacks -- sar_adc, sar_comparator_psp, mos_psp103_50c_data, worktrees/reap pass, ngram_hslm blocked; cap equals reality at 129 (Closes #10887) #10889).
• duplicate-bodies fails: ten PTX emitter helpers (at_text, code_at, count_text, emit_epilogue, emit_lds, emit_mma_step, emit_store, ins, put_quad, put_s) are copied from the forward PTX spec into specs/igla/coder/hslm_bwd_ptx.t27 (spec(igla): hslm_bwd_ptx -- HSLM backward dX and dW as int8 tensor-core PTX written by t27 (Closes #10847) #10855) and sit in no ledger. Task Bless the ledgers #10855 left red: t27b row for hslm_bwd_ptx, dupe baseline, orphan ceiling, types Range #10930 already covers this; this issue only needs that task done or the rows blessed.
• "Every source file is reachable from its crate root" fails: bootstrap: orphaned rose 1 -> 4.

Requirements

• FR-001: The scan check MUST pass on master: the test path in reap.t27 MUST stop matching the devhome pattern while testing the same rule, and the spec MUST be resealed.
• FR-002: The duplicate-bodies check MUST pass on master, by shared helpers or by blessing the ten groups in the ledger with Bless the ledgers #10855 left red: t27b row for hslm_bwd_ptx, dupe baseline, orphan ceiling, types Range #10930 named.
• FR-003: The orphan check MUST pass on master, or an issue MUST name the three newly orphaned bootstrap files and the commit that orphaned them.
• FR-004: PR igla cell 25 published: the whole-model HSLM trained by t27c --backend ml is on Hugging Face (Closes #10833) #10836 MUST merge once its required checks (validate, check-linked-issue, parse-ratchet, own-language) are green; its no-auto-merge label was removed after the review was pinned.

Success Criteria

• gh run list --branch master shows scan, duplicate-bodies and the orphan workflow concluding success on the master tip after the fix PR merges.
• t27c test-report specs/tri/worktrees/reap.t27 on the Railway lab: all pass, 0 vacuous; t27c seal --verify prints all hashes MATCH.
• gh pr view 10836 --json state returns MERGED.

Success Criteria

• gh run list --branch master --json conclusion,workflowName --jq '.[] | select(.workflowName | test("scan|duplicate-bodies|orphan")) | .conclusion' outputs success for each workflow on master tip after the fix PR merges
• t27c test-report specs/tri/worktrees/reap.t27 exits 0 with all tests passing, 0 vacuous
• t27c seal --verify specs/tri/worktrees/reap.t27 exits 0 and prints all hashes MATCH
• gh pr view 10836 --json state --jq '.state' outputs MERGED