[chenylee-aws]

Issue #, if available:

Description of changes:

With the default skipShardSyncAtWorkerInitializationIfLeasesExist=false, ProcessTask called shardDetector.shard() on every batch to get the shard's hash key range for KPL de-duplication. On a cold or stale shard cache (worker start, reshard, lease takeover) that calls ListShards on the data path, where a failure used to drop the batch.

KCL stores the hash key range on every lease since 2.3, so this change takes it from the lease instead:

• DynamoDBLeaseCoordinator.convertLeaseToAssignment copies the lease's range into a new optional ShardInfo field (new constructor overload; existing constructors unchanged; excluded from equals/hashCode/toString).
• ShardConsumerArgument.hashKeyRange() reads it from ShardInfo and caches it per consumer, falling back to the ShardDetector only for legacy leases or custom coordinators that don't set it. A ShardDetector failure propagates, so the batch is retried by the dispatch retry from Retry records delivery in shardConsumer #1822 instead of being delivered without range filtering.
• ProcessTask takes the resolved range instead of a ShardDetector.

No behavior change for skip=true (still no range filtering). Additive API only.

Testing:

• Unit: ConsumerStatesTest, ShardInfoTest, DynamoDBLeaseCoordinatorTest, ProcessTaskTest; mvn clean install passes.
• End-to-end on a 2-shard stream with KPL aggregates whose user records hash into their own shard (100 "in") or into the other shard (100 "cross", the post-reshard duplicate case). A TaskExecutionListener recorded ShardInfo.hashKeyRange() per process task and an SDK interceptor classified every ListShards caller:

Run	ShardInfo range	Delivered in / cross	Data-path ListShards
This PR, skip=false	matches the lease and shard ranges	100 / 0 (filtered)	0
This PR, skip=true	set, unused	100 / 100 (unchanged: no filtering)	0
This PR, legacy leases (range removed)	null	100 / 0	1 (fallback, then backfilled)
master, skip=false	n/a	100 / 0	1 (ProcessTask.<init>)

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.