[spec] re-save .trinity/seals/ml_host_io_model.json on Linux -- master red on spec-guards and coverage after #11094 #11097

Goal

Master is red on two checks since fcb3bc5 (#11094): spec-guards ("Does every seal still describe what the compiler emits") and coverage ("Every seal still describes its spec"). Exactly one seal drifted. Re-save it on Linux.

Why this matters

#11094 changed the Verilog emitter (no $signed(...) around an assignment target, Closes #10780). Of the 26 seals whose gen_hash_verilog moved, 25 were re-saved in that PR. The 26th, .trinity/seals/ml_host_io_model.json, cannot be saved on macOS: the host-I/O glue of specs/ml/host_io_model.t27 is Linux-only, so t27c seal --save refuses there. The PR was merged by the owner's token before a Linux run, so master carries the old hash.

Measured in the spec-guards log of run 39287408264: ml_host_io_model.json: gen_hash_verilog sealed=7ee920eef949 current=a8beaecd9847. Nothing else drifted.

Source of truth

t27c seal specs/ml/host_io_model.t27 --save on a t27c built from master. Seals are written only by that command; do not edit the JSON by hand and do not use --force (it records failures).

Deliverable

One commit changing only .trinity/seals/ml_host_io_model.json (expected: gen_hash_verilog -> sha256:a8beaecd9847..., sealed_at, built_by; tests stay 3/3, 0 vacuous).

Done when

spec-guards and coverage are green again on master.

How to verify

cargo build --release -p t27c
./target/release/t27c seal specs/ml/host_io_model.t27 --save
./target/release/t27c seal specs/ml/host_io_model.t27 --verify
git diff --stat   # one file

Risks

None beyond the seal: the Verilog change is already reviewed and merged. If --save reports a regression on Linux, stop and report the output here instead of forcing.

Links

#11094 (the emitter change), #10780 (the bug it fixes), request comment on #11094 (16:24Z).