[fix][broker] Avoid blocking the metrics thread on the pending-ack managed ledger

[merlimat]

Motivation

TransactionAggregator.generate (per-subscription transaction metrics) read the pending-ack store's managed ledger with ((PersistentSubscription) subscription).getPendingAckManageLedger().get() — an un-timed CompletableFuture.get() running on the prometheus-stats executor (a bounded, 4-thread pool).

The guard checkIfPendingAckStoreInit() only verifies pendingAckStoreFuture.isDone(); getStoreManageLedger() then composes pendingAckStoreFuture.thenCompose(store -> store.getManagedLedger()), so the managed-ledger future can still be resolving. A stuck future blocks that metrics thread, and enough of them exhaust the pool — hanging all metric scraping.

Modifications

Use getNow(null) instead of get(): if the managed ledger is already resolved it is used exactly as before; if not, this subscription's pending-ack-store stats are skipped for the current scrape and picked up on the next one. A future that completed exceptionally still throws into the existing catch (unchanged).

Verifying this change

Covered by existing tests, which pass: TransactionMetricsTest.testManagedLedgerMetrics (ledger ready → stats present) and testManagedLedgerMetricsWhenPendingAckNotInit (store not initialized → skipped).

Does this pull request potentially affect one of the following parts:

If the box was checked, please highlight the changes

• Dependencies (add or upgrade a dependency)
• The public API
• The schema
• The default values of configurations
• The threading model
• The binary protocol
• The REST endpoints
• The admin CLI options
• The metrics
• Anything that affects deployment