fix: improve NHN storage and Alibaba memory handling, retain chart 0.1.0

[eunjun-strato]

Summary

• Validate NHN Cinder CSI registration before deploying bundled persistent applications. Enable a bounded volume-permission init container only for selected Cinder volumes; keep the application non-root and preserve existing data files.
• Explicitly request ext4 for AM-created NHN StorageClasses while retaining Retain / WaitForFirstConsumer and leaving existing/default classes unchanged.
• Prefer Spider's positive finite MemSizeMib over provider-specific metadata, fixing Alibaba's actual 16 GiB capacity being interpreted as zero. Keep legacy response fallbacks and capacity checks.
• Keep all five bundled charts at 0.1.0 as requested. Do not migrate catalog versions at startup. Updated templates are supplied by the running AM image, so use its source commit/image digest to distinguish installations.
• Add regression coverage and live-test/cleanup reports. Existing custom mappings and running releases are not automatically changed.

Validation on the final source

• Full Gradle suite: 688 tests, 0 failures/errors, 1 opt-in Docker/SSH smoke test skipped.
• Helm lint/template and AM deployment-values pipeline tests included.
• Built-in frontend form: 52 scenarios plus storage/identity checks.
• Install SW integration: 18 parser cases, 7 VM-target resolution cases, iframe-to-deploy wiring.
• VM clustering: 24 catalog cases, 288 rendered visibility cases, 4 transitions, 4 submission cases.
• Frontend type check and production build passed; git diff --check passed.

Live verification and scope

• Earlier isolated Alibaba, NHN and Tencent runs exercised Apache, Tomcat, Redis, MariaDB and PostgreSQL. HTTP responses, database authentication/rejection and persistence after Pod replacement passed; owned test resources were cleaned up.
• Those live tests used the earlier test build with the proposed 0.1.1 metadata. The final decision to retain 0.1.0 was verified locally; cloud resources were not recreated for that metadata-only adjustment.
• NHN managed Cinder CSI remains a cluster prerequisite; AM does not silently install the shared add-on or relax Pod Security restrictions.
• Tencent's Redis same-Pod-to-own-Service timeout remains unresolved. Separate-client Pod access passed; this PR does not claim all networking is fixed.
• Public Ingress/CIDR, browser end-to-end, HA/multi-node and load testing were outside these live tests.
• No production server/image deployment is included in this PR creation. Local .DS_Store and handover HTML are excluded.

Reports

• doc/k8s-csp-storage-fixes-20260923.md
• doc/k8s-nhn-tencent-retest-20260928.md