Enhance GCP BYOC-I tiered, psc, booter bootstrap & etc

[wenhuiZilliz]

Summary

• support tiered， add configurable local SSD counts for GCP BYOC-I node pools
  • keep search node pool at 4 local SSDs
  • configure tiered node pool with 8 local SSDs
• switch GCP BYOC-I default resource prefix to zilliz-dp-<last-12-chars-of-data_plane_id> and remove the random provider dependency
• pass PSC endpoint IP to the booter whenever Private Service Connect is enabled, so cloud-agent pods can resolve the tunnel server through hostAliases
• reduce successful booter VM self-delete TTL from 1800s to 60s
• add optional booter serial console log streaming during terraform apply for troubleshooting
• expand GCP BYOC-I docs for resource naming, destroy flows, bucket handling, PSC behavior, and booter debugging
• wait kubectl-proxy container k8sapi ready then invoke control plane create dp api

Validation

• terraform init -backend=false
• terraform validate
• git diff --check

Notes

• booter_print_serial_logs_on_apply defaults to false; when enabled, the Terraform runner must have gcloud installed and authenticated.
• Existing deployments that used the old random resource prefix should set customer_* name variables before applying this version.

[yellow-shine]

/lgtm