[WIP] Validate and collect npd status in e2e test

[Copilot]

Adds joined-node E2E validation for node-problem-detector status and collects node-problem-detector logs during E2E log dumps.

Also updates E2E documentation to mention NPD diagnostics and runs targeted validation/security checks.

[bcho]

@copilot in e2e test, can we also dump logs from npd?

[Copilot]

@copilot in e2e test, can we also dump logs from npd?

Added NPD journal collection to the E2E log dump. Each VM now writes ${prefix}-npd.log with node-problem-detector.service logs from the active machine, with host fallback diagnostics. Addressed in d89a631.

[bcho]

Confirmed from the failed E2E run logs that this validation is flushing out the issue mentioned in #193. All Flex nodes joined and reached Ready, and node-problem-detector.service briefly reported active, but NPD was crash-looping before it could publish node conditions. The collected NPD journals repeatedly show:

panic: stat /var/lib/kubelet/kubelet/kubeconfig: no such file or directory

So the failure is caused by the bad kubeconfig path currently passed to NPD, not by node join or the detector rules themselves. Once #193 is combined with this PR, we expect the NPD condition check to pass and validate the fix.

[bcho]

included in #198