Improve pwru availability under high-volume traffic

Running pwru to capture traffic between cilium's test-conn-disrupt-server and test-conn-disrupt-client (`cilium-cli connectivity test --include-conn-disrupt-test --conn-disrupt-test-setup     --conn-disrupt-dispatch-interval 0ms`) ends up with inaccurate output. There could be multiple reasons, like bpf hash maps are full. I haven't looked into the details, open this issue for tracking.