HDDS-15679. [Recon] Fix DN dropdown mismatch with pending deletion API

[priyeshkaratha]

What changes were proposed in this pull request?

The DN pending deletion API is fetched with limit=15, returning pending block size data for at most 15 datanodes. However, the node-selector dropdown on the Capacity page was populated from storageDistribution, which returns all datanodes in the cluster. This caused a mismatch: the dropdown could list nodes for which no pending deletion data existed, while the nodes actually covered by the API response might be missing from the dropdown. Selecting such a node would show a pending block size of zero regardless of the actual value.

What is the link to the Apache JIRA

HDDS-15679

How was this patch tested?

Tested manually and using newly added unit test (unit test is generated using AI)

[ArafatKhan2198]

Thanks for the changes @priyeshkaratha
Some comments for you -

[spacemonkd]

I had a few review comments. Overall can we make this PR focus on just the dropdown issues and target the polling issues - if any, in a separate PR?

Also can we add a test for the dropdown filter — e.g., add a third DN to StorageDistribution.dataNodeUsage that is NOT in DnPendingDeletion.pendingDeletionPerDataNode, and assert it doesn't appear in the dropdown.

[priyeshkaratha]

Thanks @ArafatKhan2198 @spacemonkd for the reviews. It make sense to handle the polling issue as part of separate ticket. I have created HDDS-15782 for the same. I will work on the other review comments.