Concurrent benchmark of the real mempool's shared-state access patterns under a Leios-scale transaction load.

It opens the actual mempool (openMempoolWithoutSyncThread) over a mocked ledger interface whose forker reads inject a configurable latency to model on-disk UTxO reads. Three roles run concurrently against it, as in a node under tx-submission load:

• Adders (tx-submission clients and local clients): each submits an independent chain of transactions via the real addTx, rate-limited to a target rate.
• Readers (tx-submission servers / block forging): call the real getSnapshot (readTMVar istate) on a configurable per-peer cadence, measuring how long a read blocks.
• Syncer (the mempool sync thread): periodically advances the ledger tip and runs the real testSyncWithLedger, which revalidates the mempool through the latency-injected forker.

With the mempool holding many transactions, this reproduces the contention between revalidation, ingestion and serving that the mempool sync targets.