Add network policy for prometheus-operator in e2e tests (#2756)

The default-deny-all-traffic NetworkPolicy in olmv1-system blocks all
egress by default. The prometheus-operator pod (label
app.kubernetes.io/name: prometheus-operator) had no matching allow
policy, so it could not reach the Kubernetes API server. This caused
intermittent e2e failures with "dial tcp 10.96.0.1:443: i/o timeout"
when the operator was scheduled on the second control-plane node in
the 2-node kind cluster used by experimental-e2e.

Add an egress-allowing NetworkPolicy for the prometheus-operator pod,
matching the existing pattern used for the Prometheus server pod.

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>