fix(arbitrum): ensure unique batch RPC request IDs

[DenisCarriere]

Summary

• assign a unique transport-level ID to every request sent by the Arbitrum batch RPC helper
• restore each response to the caller-provided semantic ID before returning it
• add a regression test covering duplicate caller IDs and out-of-order responses

Why

The Robinhood Chain deployment uses an Ethereum L1 JSON-RPC endpoint that enforces the JSON-RPC 2.0 batch requirement that every request ID be unique. Several Arbitrum workers intentionally use a batch number as their semantic request ID, so multiple requests for blocks in the same batch can produce duplicate IDs.

The endpoint rejects those batches with:

JSON RPC parse error: duplicate request ID: {0}

That prevents the Arbitrum batches and L1 message trackers from populating deposits and transaction-batch data. Fixing this centrally preserves the existing caller contract while making requests valid for strict JSON-RPC servers.

Validation

• CHAIN_TYPE=arbitrum mix test apps/indexer/test/indexer/fetcher/arbitrum/utils/rpc_test.exs — 1 test, 0 failures
• mix deps.compile && mix compile — passed
• mix format — passed
• mix credo --strict on the changed files — no new findings (one pre-existing TODO suggestion remains)
• git diff --check — passed

Deployment

After merge, publish the updated Blockscout and Firehose images, update the Kubernetes image tag, and verify that the live Arbitrum batch/message trackers progress without duplicate-ID errors.