fix(discov): refresh etcd client after publisher watch errors

[014-code]

What does this PR do?

Fixes #5707.

This PR prevents the etcd publisher from repeatedly reusing a stale cached client after watch/keepalive failures. When the publisher's keepalive channel closes or the watch stream returns an error, it now revokes the current lease and invalidates the cached etcd client before re-registering, allowing a fresh client with a valid auth token to be created.

It also adds single-resource removal support to syncx.ResourceManager, used by the discovery registry to invalidate one cached etcd connection without closing the whole manager.

Tests

Added tests:

• TestResourceManager_RemoveResource
• TestPublisher_keepAliveAsyncWatchErrorRefreshesClient

Passed:

• go test ./core/discov/... -count=1
• go test ./core/syncx -run TestResourceManager_RemoveResource -count=1

Note:

• go test ./core/syncx -count=1 still fails on existing unrelated test TestImmutableResourceErrorRefreshAlways, expecting count 2 but got 1.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!