test: extend DynamoDB performance benchmarks

[iulianbudau]

Motivation and Context

AWS SDK for Java v2 already has useful DynamoDB protocol/marshalling and Enhanced Client mapper benchmarks, but coverage is fragmented across different measurement boundaries. Those existing benches remain valuable; this change complements them rather than replacing them.
This PR adds a coherent DynamoDB-specific suite under test/sdk-benchmarks so maintainers can evaluate:

• deterministic full SDK client-side performance (mocked HTTP)
• overhead across LOW, DOCUMENT, and TYPED API abstractions
• sync vs async single-operation completion
• end-to-end behavior against real DynamoDB (explicit opt-in only)
  All newly added benchmarks are manual-execution only. This PR does not modify BenchmarkRunner, MAPPER_BENCHMARKS skip behavior, shards, baseline.json, regression gates, CI workflows, or scheduled live execution.

Modifications

Tier C — Mocked full SDK pipeline

Deterministic client-side path (Enhanced mapping where applicable → endpoint resolution → marshalling → SigV4 → HTTP construction → mocked HTTP → unmarshalling → response mapping).
Coverage:

• LOW sync: GetItem, PutItem
• DOCUMENT sync: GetItem, PutItem
• TYPED sync: GetItem, PutItem, Query (first page)
• LOW async: GetItem (completion via join())
• TYPED async: GetItem (completion via join())
  Shared deterministic fixture and operation-aware canned responses via existing sync MockHttpClient and MockAsyncHttpClient.

Tier D — Live DynamoDB

Coverage: LOW GetItem/PutItem; TYPED GetItem/PutItem/Query (first page).
Measures SDK + HTTP + network + DynamoDB service latency. Safety model: explicit opt-in before credentials/client/AWS calls; unique owned PAY_PER_REQUEST tables with ownership tags; teardown deletes only the trial-created table; lightweight retry observer (opt-out available); synthetic data only.

Docs / deps

• test/sdk-benchmarks/DYNAMODB.md + README pointer (tiers, commands, live safety, interpretation)
• junit (test) and service-test-utils (compile, with unused test-only transitive exclusions) for async mock HTTP

Methodology

• Tier C: JMH AverageTime, µs, 5/5/2
• Tier D: JMH SampleTime, ms, 3/5/1
• Live LOW vs TYPED deltas are directional only (network/service dominate)

Deferred (intentionally out of scope)

CI/shard/baseline integration; async Put/Query and concurrency; broader DOCUMENT async/live; Batch/Transact; expanded parameter matrices.

Testing

• sdk-benchmarks module build + checkstyle passed
• Infrastructure and live-guard unit tests passed
• Manual Tier C LOW / DOCUMENT / TYPED and async Get runs
• Async Get under full 5/5/2 showed stable CV (~3–5%)
• Live suite with opt-in: tables created and cleaned up; no retries/throttling in that run
• Confirmed no changes to automated benchmark selection/execution paths
  Representative numbers from one validation environment only (not targets):
• Tier C sync: ~30–35 µs/op
• Tier C async Get: ~105–110 µs/op
• Tier D live: ~70–72 ms/op

Screenshots (if appropriate)

N/A

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)

Checklist

• I have read the CONTRIBUTING document
• Local run of mvn install succeeds
• My code follows the code style of this project
• My change requires a change to the Javadoc documentation
• I have updated the Javadoc documentation accordingly
• I have added tests to cover my changes
• All new and existing tests passed
• I have added a changelog entry. Adding a new entry must be accomplished by running the scripts/new-change script and following the instructions. Commit the new file created by the script in .changes/next-release with your changes.
• My change is to implement 1.11 parity feature and I have updated LaunchChangelog

License

• I confirm that this pull request can be released under the Apache 2 license