fix(cluster-tool): stop passing a reserve code on collateral deposits (SOL-432)

[valthon]

Coordinated change — SOL-432. Three identically-named branches
(fix/SOL-432-drop-deposit-reserve-code) across wire-solana, wire-ethereum
and wire-tools-ts. The deposit action's reserve_code field is caller-supplied
and unused; all three sides must land together for the e2e flow-test workflow
to be able to verify them.

Paired with: Wire-Network/wire-solana#515, Wire-Network/wire-ethereum#197

The harness threaded WireUnderwriterTool.PrimaryReserveCode into both
non-native collateral deposit steps -- SVM opp-outpost::deposit_non_native and
EVM OperatorRegistry.depositNonNative. SOL-432 (CertiK WNS-165) removes that
parameter from both outposts, so the harness stops supplying it.

The depot is the evidence that 0 is the normative value, and that a reserve code
never belonged on a deposit at all:

• wire-sysio/libraries/opp/proto/sysio/opp/attestations/attestations.proto
  declares OperatorAction.reserve_code (field 9) "SLASH-only ... Empty (0)
  for non-SLASH action types."
• wire-sysio/contracts/sysio.msgch/src/sysio.msgch.cpp dispatch_operator_action
  forwards ACTION_TYPE_DEPOSIT_REQUEST to opreg::depositinle as
  (account, chain_code, token_code, amount, op_address.kind,
  op_address.address, original_message_id) -- oa.reserve_code is not read.
• The depot's own sysio.opreg.cpp build_deposit_action leaves it
  default-initialized (0); only build_slash_action populates it.

Removal beats validation on the outposts: validating the code would require
resolving a Reserve on the collateral path, coupling the collateral custody
surface to the reserves surface. .claude/rules/outpost-three-concerns.md
permits exactly one cross-concern token edge -- the slash-to-reserve hop -- and
forbids adding another. PrimaryReserveCode had no other reader once both call
sites dropped it, so the constant is deleted; the identically-named constants in
the flow-swap-* / flow-underwriter-slashing packages are a DIFFERENT symbol
on the reserves/swap path and are untouched, as are
SolanaOutpostBootstrapper.PrimaryReserveCodename and RegistrySteps'
regreserve registration.

Per .claude/rules/no-back-compat-before-release.md the signature churn is free
pre-release, so the field is deleted from both DepositNonNativeInput shapes
rather than made optional.

Mutation -> test mapping: re-introduce reserveCode on either StepInput and the
exact toEqual assertions in EthereumCollateralTool.test.ts and the new
SolanaCollateralTool.test.ts "captures the full typed input" cases fail --
toEqual is exact, so it pins the ABSENCE of the field, not just the presence of
the others. Re-adding the positional argument to the anchor depositNonNative
call or the ethers DepositNonNativeArgs tuple is a compile error against the
regenerated IDL / new ABI. The SVM factory had NO coverage before this change;
per STYLE.md ("unit tests required for every symbol") the modified
planNonNativeDeposit / runNonNativeDeposit pair now ships the same two cases
the EVM side already had.

Lands together with the wire-solana change (deposit_non_native) and the
wire-ethereum change (OperatorRegistry.depositNonNative); this repo consumes
both signatures, so building it against un-updated outposts fails. The three are
verified as one combination by the e2e flow-test workflow and must merge
together.

Co-Authored-By: Claude Opus 5 noreply@anthropic.com
Claude-Session: https://claude.ai/code/session_01NoaXitAvpNSNj7114S9338

---

🤖 Generated with Claude Code

https://claude.ai/code/session_01NoaXitAvpNSNj7114S9338