Optimize Esplora witness resolution

[Jainakin]

Summary

• replace the blocking Esplora resolver's sequential get_tx() and get_tx_status() calls with one get_tx_info() request
• preserve confirmed, tentative, unresolved, invalid-height, and typed resolver-error behavior
• rely on the existing CheckedWitnessResolver boundary for transaction-identity validation instead of duplicating that policy in the transport adapter
• add deterministic loopback coverage for the combined endpoint and failure paths

Current head: 650370c6603d9d7f2810b91489f41444acfc867f. The branch contains the focused resolver commit plus maintainer commit 650370c fixing the current nightly Infallible conversion build. Both commits have verified signatures, and the current head is reviewer-approved.

Motivation

Each RGB witness resolution previously performed two sequential Esplora round trips: one for the raw transaction and one for its status. High-history consignments repeat this operation across validation and acceptance, making wall time proportional to both witness count and network latency.

The combined Esplora /tx/{txid} response contains both transaction and status data. This reduces each resolution from two HTTP requests to one without adding cache, concurrency, persistence, or protocol-state semantics.

Controlled 88-history result

The same 88-witness consignment was replayed through the RGB validation/import/acceptance path behind a deterministic Signet-like latency proxy:

Resolver	Full replay	HTTP requests	Unique witnesses
Existing raw + status requests	234.59 s	176	88
Combined transaction-info request	120.84 s	88	88

This is a 1.94x speedup, or a 48.5% wall-time reduction, in the controlled latency model. It is not a production Signet SLA claim. The benefit is removal of one network round trip per resolution, not lower payload bandwidth.

Correctness coverage

Deterministic tests cover:

• confirmed witnesses through one /tx/{txid} request
• unconfirmed witnesses through one /tx/{txid} request
• missing transactions returning Unresolved
• zero confirmation height returning InvalidResolverData
• malformed JSON retaining the existing typed resolver error

The reviewer-requested helper names are applied, and the redundant adapter-level txid checks were removed. The focused resolver behavior is covered by deterministic tests, and the aggregate Codecov check passes on the current head.

Validation

• focused Esplora resolver suite: 5/5 passed
• all 50 checks reported for the current head pass, including formatting, clippy, docs, no-default, feature, WASM, stable, beta, nightly, MSRV, coverage, Linux, and macOS

Scope

This PR is intentionally Esplora-only. It adds no cache, worker pool, persistent state, Electrum behavior, WASM transport, RGB acceptance change, or Lightning funding change.

• rgb-ops #18 remains the separate native Electrum batching PR.
• rgb-lib #79 remains the separate RGB-lib integration for the Electrum batch response.

This change does not parallelize witness resolution or remove repeated validation/acceptance passes. It only makes each Esplora resolution one request so subsequent parallel resolution does not retain the current two-request cost per witness.

[codecov]

Codecov Report

❌ Patch coverage is 96.03960% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/persistence/stock.rs	0.0%	4 Missing ⚠️

📢 Thoughts on this report? Let us know!

[zoedberg]

@Jainakin Thanks, nice catch! Please address the requested changes, amend them to keep a single commit, sign it with gpg and rebase it on the updated master tip (d7e96f6)

[zoedberg]

LGTM, thank you!