Add Helm resilience profile

[Osvaldo Andrade (osvaldoandrade)]

Summary

• add a Helm resilience profile that renders at least 3 database Pods, RF=3, generated StatefulSet peer maps, PDB, anti-affinity, topology spread, and longer startup probes
• add Helm validation failures for unsafe RF, replica, persistence, and memory-limit combinations
• add resource-policy support to omit database CPU limits while retaining memory limits, plus docs for production and local M1/M4 performance testing
• isolate cefas-manager pod labels from database Service selectors so manager Pods cannot receive database traffic

Validation

• go test ./...
• helm lint dist/helm/cefas
• scripts/test_helm_resilience.sh
• helm template cefas-default dist/helm/cefas --namespace cefas-test
• helm template cefas-perf dist/helm/cefas --namespace cefas-test --set resilience.enabled=true --set resourcePolicy.disableCPULimits=true
• RESULT_DIR=/tmp/cefas-bench/helm-resilience-rf3-standard PROJECT=cefas-helm-resilience-rf3 REPLICATION_FACTOR=3 scripts/bench/bench_8node_matrix.sh
  • write_only: 56,340,500 writes, 187,316/s, 0 errors
  • read_only: 32,644,134 reads, 108,807/s, 0 errors
  • mixed: 29,401,500 writes at 97,862/s and 14,956,967 reads at 49,783/s, 0 errors

Closes #573
Closes #585
Closes #586
Closes #587