Improve pwru availability under high-volume traffic

[jschwinger233]

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.