chore: annotate the intentionally-unused code paths and drop orphaned test fixtures

[blacks1ne]

49 warnings, and the most opinionated PR in the series — this is the one to
push back on if any of it reads wrong.

Almost none of these are accidental leftovers. They are ported-for-parity
mirrors of the Go implementation, accessors on library surfaces the node itself
does not call, or scaffolding for a feature that is implemented but not wired up.
So each gets #[allow(dead_code)] and a one-line note saying why, following the
annotation already on AppShardProposal::state_bytes. Scoped per item rather
than per crate, so genuinely new dead code still warns.

Two exceptions are deleted, both #[cfg(test)] fixtures with no surviving test
referencing them: token_intrinsic/pending.rs's five builders (69 lines) and
global_intrinsic/conversions.rs::sample_addr_sig.

The unreachable _ => return None in consensus_wire is annotated rather than
removed — its comment says it exists so that regenerating the protos drops an
unported variant instead of breaking the build, which seems worth keeping.

Two things I flagged but did not change

• engines::require_traversal_proof_for_inputs is dead code. Its own doc
  comment says that without this gate "an attacker can fabricate inputs whose
  values they choose and mint QUIL from nothing". The check is implemented;
  nothing calls it. I annotated it with a NOTE pointing at that rather than
  wiring it up, because wiring it up is a decision about the token dispatch path,
  not a warning fix. Flagging it here because it should not stay invisible.
• thread_worker's JoinHandle is never awaited — shutdown goes through the
  CancellationToken only.

Merge-order note — resolved

This PR and #599 used to touch adjacent lines of the same
channel::json_to_metadata signature, and would have conflicted on whichever
merged second. Rather than leave you to hit it, the whole site now lives here
(+1): the unused ratchet_state parameter is renamed in this PR alongside the
annotation and the comment, since that entire function is dead, which is what
this bucket is about. #599 keeps the rename on metadata_to_json, which is live.
The warning is still fixed exactly once across the series — it changed bucket,
it was not dropped. Verified: the two now merge in either order and produce an
identical tree.

---

Series

Part of the warning-cleanup series that starts with #597. The eight PRs are
disjoint and each stands on its own, but they are meant to be read in order —
please take #597 first: it is the only one of the eight that fixes a bug
rather than a warning, and it is the shortest.

• Base: 932045a3
• Verified with all eight applied: the workspace goes from 309 warnings to 16 —
  the 16 being classgroup's GMP FFI glue, deliberately left visible rather than
  silenced. Update: those 16 are now fixed rather than documented, in fix: correct classgroup GMP FFI declarations and uninitialized values #605,
  which corrects the declarations and the uninitialized values instead of
  annotating them. With fix: correct classgroup GMP FFI declarations and uninitialized values #605 and the channelwasm commit in chore: drop unused imports and bindings in the crypto crates #599, the combined tree
  checks with zero warnings.
• To check this PR on its own:

  cargo check --keep-going --workspace --lib --bins --tests --examples
  

  (The earlier form of this command carried --exclude channelwasm. That crate
  is in scope now — chore: drop unused imports and bindings in the crypto crates #599 covers it — so the exclusion is gone.)

Happy to re-pace these, drop any of them, or squash the set into a single PR if
you would rather review it in one pass — just say which.

Drafted with Claude Code; every site was read individually and the reasoning is
in the commit message.