Fix RBAC storage/recovery, liquidity router pool allowlist, batch payment ID collisions, dispute test coverage

[MJigah]

Fixes 4 issues assigned to me in the Stellar Program.

Changes

• rbac_access stores per-account roles in instance() storage; no admin recovery #1026: Move rbac_access role storage from instance() to persistent() (instance storage is loaded on every single contract call and doesn't scale with role count). Add SuperAdmin tracking, block self-revocation, and block removing the last SuperAdmin so the contract can't be permanently locked out.
• liquidity_router::check_and_route trusts caller-supplied pool address #1025: Add an admin-managed pool allowlist to liquidity_router (initialize/add_pool/remove_pool) and reject unapproved pool_address values in check_and_route before the cross-contract call, so an attacker-deployed contract can't fabricate reserve values.
• batch_payments::create_batch payment IDs can collide across concurrent batches #1024: Derive batch_payments payment IDs from the merchant address, a monotonically increasing per-contract counter, and the payment's own contents — instead of just ledger sequence + batch index — so IDs can't collide across merchants or concurrent batches landing in the same ledger. Also persist PaymentRecord on-chain for auditability (previously defined but never used).
• No test covers dispute() after a partial release on the same payment #1023: Document dispute()'s intended behavior when called after a partial release, and add tests covering dispute-after-partial-release resolved to both the customer and the merchant (previously untested).

Also fixed

payment_escrow/src/test.rs's setup_env() had unresolved git merge-conflict markers on main and called the contract constructor with the wrong number of arguments on both sides of the conflict — needed a working setup_env() to add the #1023 tests. Left every other pre-existing conflicted test in that file untouched, as that's unrelated to these 4 issues.

Also fixed two latent comparison bugs (Address == &Address type mismatches from misreading Vec::iter()'s item type) introduced while writing the #1025/#1026 fixes, caught before commit.

Closes #1026
Closes #1025
Closes #1024
Closes #1023

[drips-wave]

@MJigah 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