IT: wire TestTarget::Devnet for CIP-104 coupon-reassignment (Option B)

[gyorgybalazsi]

Summary

Finish the TestTarget::Devnet wiring so the CIP-104 Mode A coupon-reassignment integration test can run automatically against devnet and reproduce the manual pre-merge exercise (reassign + 0.8/0.2 split verified against real cbtc-network coupons).

This is the Option B follow-up to the localnet-seeding hardening (Option A), which lands separately and gives deterministic every-PR CI coverage. Option B is complementary: it targets the real network, so it can never be a required CI gate (it depends on live coupon timing and Mode-B coordination) — its value is an on-demand, automated reproduction of the manual devnet run.

Context: PR #255 (CIP-104 Mode A coupon-reassignment automation) is the feature; the IT hardening is being done as separate follow-up PR(s). The devnet scaffolding is already half-built (TestTarget::{Localnet,Devnet}, discover_network_parties(), Keycloak auth::Refresher, ParticipantAdminCreds).

Blockers to wire

1. Target-aware HTTP layer. crates/decman/tests/common/http.rs hardcodes http://localhost:{port} and always attaches the static/mock bearer. Make the base URL target-aware (HTTPS host on devnet) and attach the auth::Refresher (Keycloak client-credentials) token so requests authenticate against real DecMan nodes.

2. Standalone reward-only entrypoint. The reward phase currently runs at the END of the full localnet bootstrap sequence in crates/decman/tests/governance_workflows.rs. Pointed at devnet, the earlier phases would try to onboard a NEW decparty (topology writes/freeze) instead of using the pre-existing cbtc-network. Add an entry that runs ONLY the reward phase against the existing devnet decparty and skips the bootstrap phases.

3. Real beneficiary/assigner parties from config/ledger. crates/decman/tests/common/phases/coupon_reassignment.rs hardcodes beneficiaries to fixture parties (p2_member/p3_member). On devnet read the real cbtc-beneficiary / operator parties (and the active attestor assigners) from config or the ledger.

4. PQS split-by-value assertion. Add the per-beneficiary field-level check against devnet PQS pqs_cbtc: one RewardCouponV2 per beneficiary with beneficiary ∈ {cbtc-beneficiary, operator} and the 0.8/0.2 amount shares. Query archives('splice-amulet:Splice.Amulet:RewardCouponV2') grouped by payload->>'beneficiary'. Reference numbers from the manual run: cbtc-beneficiary 27,689.39 vs operator 6,922.35 (exactly 4:1).

Acceptance

• DECPM_IT_TARGET=devnet DECPM_IT_REWARD=1 <entrypoint> runs the reward phase against cbtc-network, exercises Delegation_Assign, and asserts the 0.8/0.2 split by value via PQS.
• Documented operational preconditions (Mode-B collection paused for cbtc-network, short DECPM_REWARD_AUTOMATION_INTERVAL_SECS, coupons re-accumulated unassigned).
• NOT added as a required CI check (non-deterministic; depends on live network state).

Non-goals

• Localnet coupon-seeding CI coverage — handled by the Option A follow-up PR.
• The ledger-level non-assigner negative — covered in DAML (test_non_assigner_cannot_reassign); a devnet ledger-level repro stays a manual ops step (no HTTP path to submit as an arbitrary party).

Key files

• crates/decman/tests/common/http.rs
• crates/decman/tests/common/mod.rs (TestTarget, Fixture, discover_network_parties)
• crates/decman/tests/common/auth.rs (Keycloak Refresher)
• crates/decman/tests/common/phases/coupon_reassignment.rs
• crates/decman/tests/governance_workflows.rs (gating / phase ordering)