[sarankumar1325]

Problem description

Fixes #3208.

When the connection pooler is disabled, cleanup only works on the objects the operator remembers in memory (c.ConnectionPooler[role].Deployment/Service/AuthSecret). Two things went wrong:

1. deleteConnectionPooler returned as soon as one deletion failed. If deleting the Service failed, the auth Secret was never attempted.
2. After an operator restart the new Cluster object starts with empty references. syncConnectionPooler then skips cleanup for a disabled pooler, because nothing is remembered. Leftover objects (the Service and auth Secret from the issue, or the whole pooler if it was disabled while the operator was down) stayed forever, and sync reported success.

Changes:

• When syncConnectionPooler first initialises the pooler objects of a role (first sync of the cluster, e.g. after a restart) and the pooler is not needed, the new discoverConnectionPoolerObjects looks up the Deployment, Service and <name>-u auth Secret by name. It only remembers objects that carry the operator's connection-pooler: <name> label, so unrelated objects with the same name are left alone. This is a one-time lookup per cluster object, not on every sync.
• deleteConnectionPooler now tries all three objects, forgets only the ones that are really deleted (or already gone), and returns the combined error. A failed object stays remembered and is retried on the next sync.
• The disabled-pooler branch also runs cleanup when only the auth Secret is left.

Linked issues

Fixes #3208

Checklist

• Your go code is formatted. Your IDE should do it automatically for you.
• You have updated generated code when introducing new fields to the acid.zalan.do api package. (n/a, no API changes)
• New configuration options are reflected in CRD validation, helm charts and sample manifests. (n/a)
• New functionality is covered by unit and/or e2e tests.
• You have checked existing open PRs for possible overlay and referenced them. (none found; Delete services with background propagation so their EndpointSlices are removed #3189 changes resources.go service deletion and doesn't overlap)

New unit tests in pkg/cluster/connection_pooler_test.go:

• TestConnectionPoolerCleanupAfterRestart: creates the pooler, then syncs a fresh Cluster with the pooler disabled (same old and new spec, as on startup) and checks that the Deployment, Service and auth Secret are gone for both roles. It fails on master.
• TestConnectionPoolerDeleteContinuesOnError: a fake client reactor fails the Service delete once. The auth Secret is still deleted, the Service stays remembered, and a second call removes it. It fails on master.
• TestConnectionPoolerCleanupKeepsForeignObjects: a Service with the pooler name but no pooler label is not deleted.

go test ./... (after go generate ./... for mocks), go vet ./pkg/cluster and staticcheck ./pkg/cluster all pass. I didn't run the e2e suite.

I wrote this change with the help of an AI coding assistant, and I reviewed and tested it.