feat(intent_settlement): proportional slashing (#193), bid window (#191), dispute flow (#188), multi-bond-token (#187)

[iam-mercy]

Summary

Implements four High-priority settlement-hardening issues in one change set.
They all touch the IntentState enum, slash_solver, fill_intent and the
solver bond model, so splitting them would mean four heavily-conflicting stacked
PRs; they're delivered together instead.

Issue	Feature
#193	Dynamic, proportional bond slashing
#191	Competitive bid-window solver bidding
#188	Dispute-resolution state machine (Filling → Disputed → Resolved)
#187	Multi-bond-token support with per-token accounting

closes #193
closes #191
closes #188
closes #187

---

⚠️ Base branch does not compile

main (c5da0fd) fails cargo check with 67 errors — a prior bad merge
dropped ~15 DataKey variants, ~7 Error variants and ~12 constants, left
duplicate Error discriminants, and triplicated the fill_intent transfer
block. This is pre-existing and identical on origin/main and
stellar-vortex-protocol/main.

This PR adds zero new compile errors (still exactly 67, all pre-existing and
all outside the code added here), but CI will stay red until main is
repaired separately. Because the crate can't build, the new tests could not be
executed and the wasm/proptest/clippy jobs cannot pass from this base. Every
addition here was written to be internally consistent and to match house style;
it needs a green main to be fully verified.

---

#193 — Proportional slashing

slash_solver no longer takes a flat 10 % of the bond. New helper
compute_slash_amount(bond, unfilled_amount):

exposure     = min(unfilled_amount, bond)          // same-token comparability
proportional = exposure / 10                        // 10% of what was at stake
cap          = bond * 1000 / 10_000                 // never worse than flat 10%
slash        = clamp(proportional, 1, min(cap, bond))

• Integer-only, cannot panic (all operands ≥ 0, no division by zero).
• Floor of 1 stroop preserves issue Add a minimum slash amount floor so tiny bonds round to a zero-value slash #32's guarantee.
• A 500k-USDC-bond solver failing a 10-USDC intent now loses ~1 USDC, not
  50 000. A minimally-bonded solver failing a huge intent is still capped at
  10 % of bond (and at 100 % of bond via exposure ≤ bond).
• SECURITY.md "Known Limitations" updated; solver_slashed event unchanged.

#191 — Competitive bid window

• Storage-key collision fixed: is_bid_window_enabled now reads a dedicated
  DataKey::BidWindowEnabled instead of borrowing DstAllowlistEnabled "as a
  placeholder". set_dst_allowlist_enabled and set_bid_window_enabled are now
  fully independent (regression test included).
• bid_intent(solver, intent_id, quoted_dst_amount) — reuses
  is_solver_eligible; records the bid only if strictly higher; tie-break:
  the first solver to a given amount keeps the lead.
• settle_bids(intent_id) — permissionless. Winner → Accepted with a fresh
  FILL_WINDOW and accept_intent's bookkeeping (active_intents,
  OpenIntents). No usable bid (nobody bid, or the leader's bond dropped
  below the floor since bidding) → re-opened as Open with a fresh
  INTENT_EXPIRY, matching expire_intent's permissionless-materialisation
  pattern rather than getting stuck in Bidding.
• get_best_bid view added.

#188 — Dispute-resolution state machine

Per docs/dispute-resolution-design.md. New states Filling, Disputed,
Resolved; new enum DisputeResolution { Upheld, Dismissed }; new
IntentRecord fields dispute_deadline, dispute_raised_at, resolution.

Entrypoint	Who	Effect
begin_fill	assigned solver	escrows a completing fill in the contract, opens DISPUTE_WINDOW (1 h)
dispute_fill	intent user	Filling → Disputed within the window
resolve_dispute	arbiter (set_arbiter, defaults to admin)	Upheld → full escrow to user + proportional slash; Dismissed → escrow − fee to user, no slash
release_fill	anyone	Filling + window elapsed → clean release (fee taken); Disputed + ARBITER_WINDOW (24 h) elapsed → full escrow to user, no slash, resolution == None marks the timeout

Escrowed tokens are held by the contract, never the solver; all state is
committed before any transfer (CEI). Every new error path has a
uniquely-discriminated Error variant (30–39).

#187 — Multi-bond-token support

Per docs/60-multi-bond-token-design.md, implemented additively (the design
doc's "remove bond_amount" schema would break every existing test and the
bond-conservation proptest with no way to verify the refactor against a
non-compiling base):

• DataKey::SolverBond(solver, token) holds per-token balances.
  SolverRecord.bond_amount is kept as the default-token mirror so pre-[High] Implement multi-bond-token support with per-token accounting #187
  readers keep working; SolverRecord.bond_tokens: Vec<Address> enumerates the
  rest (cap MAX_BOND_TOKENS = 8).
• register_solver / withdraw_bond / accept_intent keep their signatures
  (pinned to the default token) and gain register_solver_with_token /
  withdraw_bond_token / accept_intent_with_bond siblings ("Option A" from
  the design doc §7.2).
• add_allowed_bond_token / remove_allowed_bond_token / set_bond_token_min /
  get_bond_token_min / get_solver_bond / get_solver_bonds.
• IntentRecord.bond_token records which token backs each accepted intent;
  slash_solver slashes and pays out in that token. deregister_solver
  refunds every token the solver holds.
• Per-token minimums: DataKey::MinBond(token), falling back to
  ProtocolConfig.min_bond for the default token and MIN_BOND otherwise.
• New errors BondTokenNotAllowed = 40, TooManyBondTokens = 41 (discriminants
  differ from the design doc's §6 to avoid colliding with existing variants).
• CHANGELOG.md [Unreleased] notes the storage-layout change and its
  upgrade behaviour.

Also

fill_intent's output/fee transfer block had been written three times by a
bad merge (paying the user 3× and the fee 2×). De-duplicated to a single
transfer after all state is committed.

Tests

25 new cases in test.rs:

• [High] Implement dynamic, proportional bond slashing #193 — proportional slash vs. huge bond / tiny intent, cap at 10 % of bond,
  floor of 1.
• [High] Implement the bid-window competitive solver bidding feature #191 — no-bid timeout re-opens as Open, single-bid settlement,
  strictly-higher replacement, set_dst_allowlist_enabled no longer touches
  bid-window state.
• [High] Implement the dispute-resolution state machine (Filling → Disputed → Resolved) #188 — happy path (release_fill), disputed→upheld (full refund + slash),
  disputed→dismissed (refund − fee, no slash), arbiter timeout, escrow custody.
• [High] Implement multi-bond-token support with per-token accounting #187 — multi-token registration, per-token slash isolation, deregister
  refunds all, unapproved-token rejection, withdraw-below-minimum, token cap,
  legacy single-token path unchanged.

proptest_bond.rs's AcceptAndSlash step now asserts the exact proportional
formula instead of the flat 10 %.

[drips-wave]

@iam-mercy Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits