Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@pkg/controller/add_hac.go`:
- Around line 11-14: The PR flips HAC_ENABLED semantics causing unset → enabled;
update documentation, release notes, and runtime logging to avoid surprise:
modify doc/high-availability-controller.md to describe the new opt-out behavior
and explicitly state that operators must set HAC_ENABLED=false to disable HAC;
add an upgrade/release note describing the behavioral change and migration steps
for clusters that relied on the previous default; and add a startup log line in
the manager initialization where AddToManagerFuncs is appended (referencing
AddToManagerFuncs and hac.Add) that logs the observed HAC_ENABLED value and
whether the HAC controller was enabled so operators can easily see the active
setting on boot.

In `@pkg/hac/hac.go`:
- Around line 52-56: The warning log and pod-filtering are fragile: replace the
prefix check strings.HasPrefix(pod.Name, "noobaa-db-pg-") with a label-based
check that reads the pod's labels (e.g., pod.Labels["noobaa-db"] == "postgres")
and skip pods matching that label, and update the log.Warningln call to include
the actual pod name (pod.Name) so the message identifies which pod was skipped;
change the filter logic and the log site where strings.HasPrefix and pod.Name
are used to reference pod.Labels and include pod.Name in the warning text.

---

Nitpick comments:
In `@pkg/hac/hac.go`:
- Around line 57-60: The current loop returns early on the first failure from
pd.Client.Delete which stops attempting deletion of remaining pods; change the
loop that calls pd.Client.Delete (referencing pod, deleteOpts, pd.Client and
pd.NodeName) to instead log each per-pod failure and append the error to a slice
(or use k8s.io/apimachinery/pkg/util/errors utilerrors.NewAggregate) so the loop
continues trying all pods, and after the loop return either nil if no errors or
the aggregated error; ensure logging still includes pod.Namespace, pod.Name and
pd.NodeName for each failure.

Verify each finding against the current code and only fix it if needed.

Duplicate comments:
In `@pkg/hac/hac.go`:
- Around line 52-56: Replace the fragile prefix check that uses
strings.HasPrefix(pod.Name, "noobaa-db-pg-") with a label-based check on
pod.Labels (e.g., if v, ok := pod.Labels["noobaa-db"]; ok && v == "postgres" {
... }) and ensure you nil-check the Labels map before reading; also change the
log call from log.Infof to the appropriate warning-level logger (e.g.,
log.Warningf or log.Warningln) while keeping the pod.Name in the message so
skipping DB pods is clearly visible to operators.