fix(valkey): add dynamic parameter reconfiguration via nc-based shellTrigger

[weicao]

Problem

The Valkey addon ParametersDefinition was missing dynamicParameters and reloadAction, causing all parameter changes (including dynamic ones like maxmemory-policy) to trigger a rolling pod restart instead of hot-reloading via CONFIG SET.

Root cause chain

1. No dynamicParameters list → KB classified all params as static → rolling restart
2. No reloadAction → even with dynamicParameters, KB had no way to apply changes
3. shellTrigger runs in config-manager sidecar (not the valkey container), so valkey-cli is unavailable (glibc/musl ABI mismatch)
4. batchReload: true passes params via argv ($0=key, $1=value), not stdin
5. Solution: send inline Redis protocol (AUTH + CONFIG SET) via nc 127.0.0.1 6379

Changes

• addons/valkey/templates/paramsdef.yaml: add dynamicParameters (51 params supporting CONFIG SET) and staticParameters, and add reloadAction.shellTrigger using nc
• addons/valkey/templates/cmpd.yaml: fix externalManaged comment

Verification

• R08a dynamic reconfigure: 14 PASS / 0 FAIL — maxmemory-policy applied to 3 pods, no restart
• R08b static restart: 22 PASS / 0 FAIL — databases=20 triggered rolling restart on all 3 pods

Known independent issue

kbcli cluster configure has a resolveConfigTemplate bug in KB 1.0.x. Tests used direct kubectl apply OpsRequest as workaround. Tracked separately.