Rewire minActiveReplica to IdealState and delete MIN_ACTIVE_REPLICAS from ResourceConfig

[sjainit]

Summary

Part of the ResourceConfig config-surface simplification epic (CICP-51496). Makes IdealState the single source of truth for minActiveReplica, physically deleting the duplicated MIN_ACTIVE_REPLICAS key from ResourceConfig.

MIN_ACTIVE_REPLICAS was never set on ResourceConfig in production (0/780 clusters set it on RC; all set it on IdealState). The only RC-side reader was DelayedRebalanceUtil.getMinActiveReplica(ResourceConfig, IdealState, int) — a 3-tier priority helper (RC → IdealState → replicaCount). Every one of its 4 callers built a merged RC inline purely to feed this call, and a merged RC already copies IdealState's value when RC is unset. So dropping the RC tier and reading IdealState directly is behavior-preserving for every real cluster.

Changes

• ResourceConfig: remove the MIN_ACTIVE_REPLICAS enum constant, both constructor params, the field-write, getMinActiveReplica(), the Builder field + getter/setter, the build() arg, and the merge line in mergeIdealStateWithResourceConfig.
• DelayedRebalanceUtil.getMinActiveReplica: drop the ResourceConfig tier; new signature getMinActiveReplica(IdealState, int) reads minActiveReplicas directly off IdealState, falling back to replicaCount. Also removed the stale TODO: Remove the idealState input once we have all the config information migrated to the resource config — that predates the campaign's IdealState-as-source-of-truth direction.
• Rewire the 4 callers (2 internal in DelayedRebalanceUtil, WagedRebalancer, DelayedAutoRebalancer) to the 2-arg helper, dropping the inline mergeIdealStateWithResourceConfig.
• ResourceConfigTrimmer: remove the MIN_ACTIVE_REPLICAS non-trimmable entry (change detection trims raw persisted RCs, which never carry it; IdealState changes are detected separately via IdealStateTrimmer).
• Tests: drop RC-side minActiveReplica assertions/setters in TestResourceConfig and TestClusterModelProvider; IdealState still sets it (setMinActiveReplicas), so merged-config assertions still exercise the value.

Testing Done

• JDK 11: mvn -o -pl helix-rest -am test -Dtest=TestResourceConfig,TestClusterModelProvider,TestHelixPropoertyTimmer,TestWagedRebalancer -Dsurefire.failIfNoSpecifiedTests=false — passes, including core and REST main/test compilation and all 38 selected test cases.
• Constructor and IdealState merge coverage preserve the other ResourceConfig settings while removing only MIN_ACTIVE_REPLICAS.

Scope and related PRs

• Standalone on dev, which already includes merged Retire legacy WeightAwareRebalanceUtil, remove unread STATE_MODEL_DEF_REF/REPLICAS from ResourceConfig #270. Only MIN_ACTIVE_REPLICAS is removed and rewired here. NUM_PARTITIONS, MAX_PARTITIONS_PER_INSTANCE, HELIX_ENABLED, and EXTERNAL_VIEW_DISABLED remain available. This PR does not include or depend on the unmerged changes in Delete dead ResourceConfig HELIX_ENABLED and EXTERNAL_VIEW_DISABLED #268, Delete unread NUM_PARTITIONS from ResourceConfig #271, or Rewire maxPartitionsPerInstance to IdealState and delete MAX_PARTITIONS_PER_INSTANCE from ResourceConfig #273.
• Deprecate ResourceConfig.Builder structural setters that duplicate IdealState #262 (deprecates RC setters) will need a rebase after this merges to drop the deleted setMinActiveReplica.

[LZD-PratyushBhatt]

wrong base branch, switch to dev. rebase the PR.

[LZD-PratyushBhatt]

wait, the asserts dropped just above here sit under the // Test priority, Resource Config field has higher priority. comment. mergeIdealStateWithResourceConfig uses setIntFieldIfAbsent, so a value set on the ResourceConfig wins and reading IdealState directly flips that. does the data scraping also cover RC znodes already persisted with the key? bcause that can change the behavior if this gets deployed

[LZD-PratyushBhatt]

just to confirm, description says 0/780 clusters, but minActiveReplica is a per resource setting and #267 counted 40,613 resources. did the scan walk every resource config inside those clusters?