What this PR does / why we need it:

• Add a scaling field to the FeatureStore CRD that supports static replica counts and HPA-based autoscaling for the Feast deployment
• Validate that file-based persistence (SQLite, DuckDB, registry.db) is rejected when horizontal scaling is configured, since these backends don't support concurrent access
• Automatically switch deployment strategy from Recreate to RollingUpdate when scaling is enabled to ensure zero-downtime scaling events

CRD & API

• Added ScalingConfig (with replicas and autoscaling fields) to FeatureStoreServices
• Added AutoscalingConfig with minReplicas, maxReplicas, metrics, and behavior (maps to HPA v2 spec)
• Added ScalingStatus to FeatureStoreStatus reporting currentReplicas and desiredReplicas
• CEL validation rule enforces mutual exclusivity of replicas and autoscaling

Controller

• New scaling.go with validation, HPA lifecycle management (create/update/delete), replica calculation, and status updates
• Helper functions isScalingEnabled() and isFilePersistence() in util.go
• RBAC for horizontalpodautoscalers (get, list, watch, create, update, delete)
• Controller watches owned HPA resources via Owns()
• Deployment strategy defaults to RollingUpdate when scaling is enabled

Documentation

• New "Horizontal Scaling with the Feast Operator" section in docs/how-to-guides/scaling-feast.md with static replicas, HPA, and KEDA examples
• Updated docs/how-to-guides/feast-on-kubernetes.md with scaling references
• Regenerated CRD API reference (ref.md) with new types
• Sample CRs: v1_featurestore_scaling_static.yaml and v1_featurestore_scaling_hpa.yaml