HDDS-16255. Avoid the per-group list copy in ContainerMapper.parseOmDB

[shuan1026]

What changes were proposed in this pull request?

ContainerMapper.parseOmDB called OmKeyLocationInfoGroup.createLocationList() once per version group of every key, only to iterate the result once. createLocationList() is documented as not O(1).

This PR adds one nested loop over the zero-copy getLocationLists() accessor; the inner loop body that builds BlockIdDetails is unchanged.

Additionally, a throwaway (not committed) JUnit micro-benchmark measured per-traversal thread allocation for both implementations (JDK 21.0.12, 50,000-iteration warmup then 500,000 measured iterations per shape, deterministic checksums asserted equal, volatile sink to block JIT elimination). Results, reproduced across two independent JVM processes:

Key Shape	Before	After	Saved
1 group x 1 block	100 ns / 419 B	8 ns / 0 B	419 B
1 group x 4 blocks	87 ns / 408 B	13 ns / 0 B	408 B
1 group x 32 blocks	242 ns / 744 B	61 ns / 0 B	744 B
4 groups x 4 blocks	242 ns / 1632 B	52 ns / 0 B	1632 B
16 groups x 8 blocks	1303 ns / 6528 B	312 ns / 0 B	6528 B

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-16255

How was this patch tested?

The following test passed:

• TestContainerMapper#testContainerMapper

Also ran checkstyle.sh and author.sh successfully.

[sravani-revuri]

Thanks @shuan1026 for the patch. few minor comments below.

[sravani-revuri]

nit: return type here says BlockDetails but the actual type is BlockIdDetails.

[sravani-revuri]

nit: this method now has 4 levels of looping which can be hard to read and understand. would it be better to put the logic below into a separate function?