CNS-145 - Flatten data for Cluster Usage table

[jdonelson]

Motivation

Feedback from Product team and others indicated that the previous design, grouped rows that separated out clusters and their replicas, should be simplified. This has an added benefit of reducing complexity in the UI code, which should make it easier to build out forthcoming enhancements.

Description

Simplify Cluster Replica Usage table by flattening the data to one row per replica, with an added cluster name column. A cluster with 2 replica now displays 2 rows. A cluster with 0 replicas in displayed with dashes in most columns.

Verification

Revised automated tests to handle single rows. Verified behavior locally. This version of the table is still gated behind a feature flag, so this PR does not touch any production-visible code.

[linear-code]

CNS-145

[def-]

QA LLM Review

1. MEDIUM -- Give the flattened rows stable IDs

console/src/platform/clusters/ClusterUsageTable.tsx:251

The flattened table relies on position-based row IDs while every replica row owns stateful Alter and Drop dialogs. A live replica insertion or removal can therefore reuse an open dialog component for a different cluster, so submitting the dialog can alter the wrong cluster.

Details

No getRowId is passed to useUniversalTable, so TanStack derives each top-level row.id from its index. UniversalTable then uses that ID as the React key. For example, if a replica is removed from cluster A while an Alter dialog is open on a later row for cluster B, the flattened indices shift and the keyed component that held B's open dialog can be reused for cluster C. AlterClusterModal keeps its original form state but reads cluster.name from the updated prop when it submits, which can apply B's requested size, replication factor, and rename to C. Replica changes did not move the cluster parent rows in the grouped table, but they do move these new top-level rows. Pass a stable getRowId, using a composite such as the cluster ID plus the replica ID and a distinct suffix for the empty-cluster row.