plugin-aws: kill()/stop() not implemented on killable/pollable tasks

Sub-issue of kestra-io/kestra-ee#10964.

**Tasks:** io.kestra.plugin.aws.kinesis.Consume, io.kestra.plugin.aws.sqs.Consume

**Why:** Long-lived stream consume loops (Kinesis shard iterator / SQS long-poll) with no kill()/stop() override — killing the execution doesn't stop the poll loop or release the SQS visibility timeout / Kinesis iterator early.

**Fix:** override `kill()` and/or `stop()` on these task classes to cancel the poll loop / close the consumer / cancel the remote job. See `io.kestra.plugin.gcp.bigquery.AbstractBigquery` for a reference implementation.

Start by reading io.kestra.plugin.gcp.bigquery.AbstractBigquery for the reference lifecycle implementation, then inspect io.kestra.plugin.aws.kinesis.Consume and io.kestra.plugin.aws.sqs.Consume. Trace their long-lived consume loops and determine how kill() or stop() should cancel polling and close the consumer. Done means killing or stopping either task ends polling and releases the relevant AWS resources.

Written by the indexing model from the issue text.