HDDS-16205. Avoid the duplicate list copy in S3MultipartUploadCompleteRequest.getMultipartDataSize

[shuan1026]

What changes were proposed in this pull request?

S3MultipartUploadCompleteRequest.getMultipartDataSize calls OmKeyLocationInfoGroup.getLocationList() twice for the same version group per MPU part: once to set partNumber on each block, and again to addAll those same blocks into partLocationInfos. getLocationList() is documented as expensive: it flatten-copies locationVersionMap.values() into a brand-new ArrayList on every call. Since OmKeyLocationInfo is a mutable object, the second call re-flattens the same data the first call already produced, just to read it once more.

Replaces both calls with a single getLocationLists() walk of the live, uncopied Collection<List<OmKeyLocationInfo>> view of locationVersionMap.values(), setting partNumber and doing addAll in one pass per inner list. Zero list copies instead of two, same forEach idiom as before.

Same accessor-usage fix already applied in KeyManagerImpl.refreshPipeline (HDDS-5384), ContainerEndpoint.getBlocks (HDDS-16202), and ContainerToKeyMapping.getContainers (HDDS-16204). Behavior and ordering are unchanged.

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-16205

How was this patch tested?

The following test suites passed (16 tests):

• TestS3MultipartUploadCompleteRequest
• TestS3MultipartUploadCompleteRequestWithFSO

Also ran checkstyle.sh and author.sh successfully.

[rich7420]

@shuan1026 thanks for the patch!
I think you could paste this in your pr content https://github.com/shuan1026/ozone/actions/runs/32261921025